ArticlePDF Available

QCheck: A Web Server for Quality Check of Next Generation Sequencing Data with Standards and Guidelines



Content may be subject to copyright.
S1 The Journal of Molecular Diagnostics
The Journal of Molecular Diagnostics, Vol. 21, Supplement, May 2019
Abstracts of the 2nd Global Congress on Molecular Pathology:
Evidence-based Precision Medicine
The Association for Molecular Pathology
May 16 - 18, 2019
Hong Kong
Genetics/Inherited Conditions
OR01. An International Interlaboratory Study of
Complex Variant Detection by NGS
S. Lincoln1, S. Mahamdallie2, A. Fellowes3, S.
Chowdhury4, E. Klee5, J. Zook6, R. Truty1, R. Garlick7, M.
Salit8, N. Rahman2, S. Kingsmore4, R. Nussbaum1, M.
Ferber5, B. Shirts9
1Invitae, San Francisco, CA, United States, 2Institute of
Cancer Research, London, United Kingdom, 3Peter
MacCallum Cancer Centre, Melbourne, Australia, 4Rady
Children's Institute for Genomic Medicine, San Diego,
CA, United States, 5Mayo Clinic, Rochester, MN, United
States, 6National Institute of Standards and Technology,
Gaithersburg, MD, United States, 7SeraCare Life
Sciences, Gaithersburg, MD, United States, 8Stanford
University, Palo Alto, CA, United States, 9University of
Washington, Seattle, WA, United States
Introduction: Next-generation sequencing (NGS) is a
capable technique for detecting single nucleotide variants
(SNVs) and small insertions and deletions (indels) in
relatively accessible parts of a patient's genome.
However, conventional NGS methods have important
limitations. An analysis of over 80,000 patients, tested for
constitutional alterations using sensitive methods,
showed that variants of other, technically challenging
types comprise between 9 and 19% of the reportable
pathogenic findings, depending on clinical indication.
Approximately 50% of these variants were of challenging
types (large indels, single exon copy number variants
(CNVs), etc.), 20% were in challenging genomic regions
(homopolymers, non-unique sequences, etc.), and 15%
were in regions poorly covered by standard commercial
kits. A further 15% presented multiple challenges. This
study has since been expanded to include over 200,000
patients with diverse indications, and these newer data
lead to similar conclusions.
It can be difficult for both clinicians and laboratory
directors to evaluate the sensitivity of tests for such
challenging variants. The most recent AMP/CAP
guidelines for NGS bioinformatics [1] recommend that a
total of 59 variants of each type be included in validation
studies, a number that is difficult to achieve for complex
variants given the relative scarcity of positive control
Methods: We developed a synthetic specimen
containing 22 challenging variants of diverse types in 7
commonly tested genes. Raw NGS data for these
synthetic variants was found to mimic that of the
endogenous variants and presented similar technical
challenges for NGS. This specimen was sequenced
using 10 different NGS tests by an international group of
collaborating laboratories. These tests employed different
sequencing platforms, library methods, and
bioinformatics pipelines.
Results: All 10 tests detected all of the relatively “easy”
SNVs and small indels present (with one exception).
However, only 10 of the 22 challenging variants were
detected by all tests, and just 3 tests detected all 22.
Limitations with large indels, homopolymer associated
variants, variants in non-unique regions, and other
challenges were observed. Many, but not all of these
limitations appeared to be bioinformatic in nature. Some
of these limitations were not previously known to the
respective laboratory directors, demonstrating the utility
of this approach.
Conclusions: We believe that both our prevalence data
and control specimens such as ours may be a valuable
asset to the global AMP community for evaluating and
optimizing clinical tests. The particular specimen
described is now available to members of the AMP
community and we are currently expanding this specimen
and others to include additional variants. We are
soliciting input from the community on additional variant
types and genes to include in new specimens.
OR02. High-Risk Cytogenetics in Multiple Myeloma:
Further Scrutiny of Deletions within the IGH Gene
Region Enhances Risk Stratification
J.N. Sanmann1, P.A. Althof1, S.C. Smith1, B.J. Dave1
1Munroe-Meyer Institute at the University of Nebraska
Medical Center, Human Genetics Laboratory, Omaha,
NE, United States
Introduction: High-risk (HR) cytogenetics in multiple
myeloma are best characterized using fluorescence in
situ hybridization (FISH), as low proliferation indices and
cryptic genetic aberrations often preclude informative
results by standard karyotyping. When combined with the
International Staging System, lactodehydrogenase, and
other clinical features, risk can be stratified and treatment
S2 The Journal of Molecular Diagnostics
options can be informed by the cytogenetics profile.
Certain translocations [i.e., t(4;14), t(14;16), t(14;20)]
involving disruption of IGH are in the HR category, while
the other IGH-disrupting translocations [i.e., t(11;14),
t(6;14)] confer standard-risk (SR). To enhance efficiency,
minimize cost, and preserve limited material (CD138+
fraction), laboratories typically use a breakapart probe
(BAP) to detect rearrangement of IGH, followed by dual
fusion probes to determine the IGH fusion partner.
However, there is a paucity of data describing
translocations and resultant fusion partners when the
BAP for IGH reveals a deletion of part or all of the 5'
(telomeric) or 3' (centromeric) gene region.
Methods: Our laboratory performed a retrospective
review of abnormal multiple myeloma FISH studies
between July 2012 and December 2017 to identify cases
harboring deletions of the IGH region using a BAP. We
catalogued these findings in conjunction with the dual
fusion studies performed [2012 - 2016: t(4;14), t(14;16),
t(11;14); 2017: the aforementioned probes plus t(14;20)]
to determine whether these IGH deletions were indicative
of clinically-significant translocations associated with HR
or SR.
Results: Of our 233 patient cohort exhibiting IGH
deletions, we identified 73 5' deletions, 29 3' deletions,
124 partial 5' deletions, and 7 partial 5' deletions with
concurrent 3' deletions. Subsequent dual fusion analysis
revealed a fusion partner in 19% (n=44) of cases. Of
these fusion-positive (FP) cases, 34% (n=15) exhibited
HR translocations, including 26% (n=9) of FP cases with
5' deletions, 100% (n=6) of FP cases with 3' deletions,
and 66% (n=2) of FP cases with partial 5' deletions. Of
note, patients with 5' deletions (partial or full) exhibited
both HR and SR translocations, while patients with 3'
deletions exhibited only HR translocations [i.e., t(4;14)].
Conclusions: Differentiating between IGH translocations
that confer HR versus SR is imperative for
prognostication and management. Although an IGH BAP
strategy is commonly used to screen cases, our data
suggest that subsequent analysis with dual fusion probe
strategies should not be restricted to cases with
traditional IGH BAP rearrangements. Specifically, our
data demonstrate that atypical signal patterns (e.g.,
deletions) are often indicative of clinically-significant IGH-
partner gene fusions. Thus, we recommend performing
dual fusion FISH testing among cases with deletion of
IGH, particularly for assessment of IGH translocations
associated with HR gene partners, to ensure appropriate
stratification and treatment.
OR03. The Clinical Utility of Tumour Mutational
Signatures for Identifying Hereditary Colorectal
Cancer and Polyposis Syndromes
D. Buchanan1,2, P. Georgeson1, M. Clendenning1, K.
Mahmood1, A. Ragunathan1,3, R. Walker1, J. Joo1,2, H.
Jayasekara1,4, C. Rosty1,5,6, F. Macrae7, B. Pope1, M.
Jenkins8, I. Winship3
1The University of Melbourne, Department of Clinical
Pathology, Melbourne, Australia, 2The University of
Melbourne, Centre for Cancer Research, Melbourne,
Australia, 3Royal Melbourne Hospital, Genomic Medicine
and Family Cancer Clinic, Parkville, Australia, 4Cancer
Council Victoria, Cancer Epidemiology and Intelligent
Division, Melbourne, Australia, 5Envoi Pathology,
Brisbane, Australia, 6The University of Queensland,
Brisbane, Australia, 7The Royal Melbourne Hospital,
Colorectal Medicine and Genetics, Parkville, Australia,
8The University of Melbourne, Centre for Epidemiology
and Biostatistics, Carlton, Australia
Introduction: Tumour mutational signatures are derived
from the type of nucleotide substitution and their
sequence context. Mutational signatures with known
aetiology include: A ubiquitous signature associated with
spontaneous deamination of 5-methylcytosine and
correlated with age (signature 1), and signatures related
to defective DNA mismatch repair (MMR; signatures 6,
15, 20, 26), polymerase proofreading exonuclease
domain mutations in POLE (signatures 10 and 14) and
base excision repair (BER) defects in NTHL1 (signature
30) and MUTYH (signature 18). The aim of this study
was to assess the utility of tumour mutational signatures
in colorectal cancers (CRCs) for identifying carriers of
pathogenic variants and classifying variants of uncertain
clinical significance (VUS) in genes underlying hereditary
colorectal and polyposis syndromes.
Methods: Hereditary CRC and polyposis susceptibility
gene mutation and VUS carriers were identified through
germline testing of participants from the Australasian
Colorectal Cancer Family Registry (MLH1, MUTYH,
POLE, POLD1, AXIN2 variant carriers) and from the
Genetics of Colonic Polyposis Study (NTHL1 and RNF43
variant carriers). Whole exome sequencing of formalin-
fixed, paraffin-embedded (FFPE) tumour DNA was
performed using Agilent Clinical Research Exome V2
sequenced on Illumina Novaseq. Using the mutational
signatures established by the Wellcome Trust Sanger
Institute, we calculated the optimal combination of
mutational signatures for each tumour sample based on
cosine similarity.
Results: Two CRCs from MLH1 gene mutation carriers
(Lynch syndrome) demonstrated signatures 6 and 15
correlating with defective MMR. The patterns in these two
carriers were not ostensibly different to a sporadic CRC
with biallelic somatic mutations in MLH1. Three CRCs
from homozygous carriers of the MUTYH (p.Gly396Asp)
mutation demonstrated a dominant signature 18
associated with defective BER. A CRC from a
heterozygous carrier of the MUTYH (p.Gly396Asp)
mutation and a VUS in MUTYH (p.Arg304Ser) did not
demonstrate signature 18 suggesting the VUS is not
pathogenic. The CRC from a biallelic NTHL1 carrier
S3 The Journal of Molecular Diagnostics
showed a dominant signature 30. A MMR-proficient CRC
from a carrier of a POLD1 VUS in the exonuclease
domain (p.Gly321AlafsTer72) showed no evidence of
hypermutation or defective DNA repair signatures. CRCs
from a truncating RNF43 (p.Arg132Ter) mutation carrier
with Serrated Polyposis Syndrome, and an AXIN2
truncating variant carrier (p.P350LfsX13) also did not
demonstrate DNA repair signatures.
Conclusions: The tumour mutational signatures
observed in CRCs from known pathogenic mutation
carriers were consistent with the underlying DNA repair
defect. The presence or absence of expected tumour
mutational DNA repair signatures in CRCs from
individuals with a VUS may help classify rare variants in
hereditary CRC and polyposis genes. Further exploration
of somatic mutation patterns in CRCs from RNF43 and
AXIN2 carriers may identify novel mutational signatures.
OR04. Prospective Study to Determine the Spectrum
of Mutations among Patients with Multiple Endocrine
Neoplasia 1 (MEN-1) and to Identify “At Risk”, First-
Degree Relatives
R. Pai1, A. Hesaraghatta2, S. Rajaratnam2, N. Thomas2,
T. Paul2, M.J. Paul3, D. Abraham3, A. Chacko4
1Christian Medical College, Pathology, Vellore, India,
2Christian Medical College, Endocrinology and
Metabolism, Vellore, India, 3Christian Medical College,
Endocrine Surgery, Vellore, India, 4Christian Medical
College, Neurosurgical Sciences, Vellore, India
Introduction: Multiple Endocrine Neoplasia (MEN-1) is
an autosomal dominant disorder characterized by the
combined occurrence of tumors of the parathyroid
glands, pancreatic islets and the anterior pituitary. About
90% of patients with familial MEN-1 disease are known to
carry mutations in the MEN1 gene. However, there is
very little information about the spectrum of mutations
seen in India.
Methods: Twenty-five clinically suspected cases of
MEN-1 were enrolled prospectively over a two-year
period. Of the 25 clinically suspected cases of MEN-1, 18
patients had definite clinical MEN-1 based on the
presence of at least 2 of the endocrine tumours
associated with MEN-1; 12 had family history of MEN-1.
Further, 16 first-degree relatives of index cases with
MEN-1 were screened for mutations that were seen in
the index cases. The MEN1 gene (exons 2-10), 3' and 5'
UTR regions of MEN1 gene, CaSR genes (exons 2-7)
and the CDKN1B gene (exons 1-2) were screened in all
index cases by PCR followed by Sanger sequencing. In
the absence of mutations, the samples were screened for
large deletions by Multiplex Ligase Dependent Probe
Amplification (MLPA) assay.
Results: Among the 25 index cases of MEN-1, 17 (68%)
had pituitary adenoma, 19 (76%) had primary
hyperparathyroidism and 17 (68%) had pancreatic
neuroendocrine tumors. Three patients had thymic
tumors, four had adrenal tumors, four had collagenomas,
one had angiofibroma and one had papillary
microcarcinoma thyroid. Of the 18 cases with definitive
MEN-1, 61% (n=11) were positive for mutations.
However, there were no hotspot of mutations, though 4 of
the 11 harboured an exon 10 mutation. None of the 7
probable cases of MEN-1 were positive for mutations.
Further, all the mutation negative cases that were
screened for large deletions by MLPA, 3' and 5' UTR
regions of MEN1 gene, CaSR and CDKN1B gene,
continued to be negative for mutations. However, certain
established polymorphisms in the MEN1 gene, CaSR
and the CDKN1B gene were seen among the mutated
and mutation negative cases. Interestingly, 7 of the 9
definitive cases of MEN-1 that were MEN1 mutation
negative were found to be positive for the p.V109G
polymorphism in the CDKN1B gene.
Sixteen first degree relatives including 9 members of a
single large family, were screened for the same
mutations as seen in the index case. Of the 9 members 6
were found to harbor the same mutations as in the index
case (c.1151_1152insGAGG; p.A384fs) while three were
negative. Among the other “at risk” families tested 3 of
the remaining 7 tested, harboured the same mutations as
in the index case.
Conclusion: The study has helped to clearly document
the spectrum of mutations seen among patients with
MEN-1. However, the absence of MEN-1 mutation in ~
39% of the cases and the presence of the p.V109G
polymorphism in the CDKN1B gene among the MEN-1
mutation-negative cases raises the question whether
such polymorphisms could be independently contributing
to pathogenesis.
OR05. Five Year Experience of Clinical Next-
Generation Sequencing for Somatic Overgrowth and
Related Syndromes: Non-PIK3CA Cumulative
Y. Cao1, M. Evenson1, M. Corliss1, M. Schroeder1, J.
Neidich1, J. Heusel1
1Washington University School of Medicine, Department
of Pathology and Immunology, St. Louis, MO, United
Introduction: A number of oncogenes with recurrent
mutations in cancer are now understood to drive
segmental overgrowth and related syndromes. Our
clinical laboratory has pioneered the sensitive detection
of causative mutations in cancer and applied that
expertise to mutations associated with overgrowth and
related phenotypes. Presented here are our non-PIK3CA
cumulative findings.
Methods: We use targeted hybridization capture coupled
with Next-Generation Sequencing (NGS) for deep and
comprehensive coverage. This assay has undergone
frequent modifications due to its growing knowledge
base. The current assay includes baits targeting exonic
regions and selected intronic sequences from 177 genes
recurrently mutated across all cancer types
(Comprehensive Cancer Gene Set, version 3.2). The 35
genes reported for somatic overgrowth are divided into
seven discrete orderable subsets depending on the
presenting phenotype. Paired-end 101bp sequencing
was performed on Illumina HiSeq 2500 with an average
unique on-target coverage depth of >1000x. Variants
S4 The Journal of Molecular Diagnostics
were called and analyzed using a combination of multiple
software programs, and Clinical Genomics Workspace
was used for data visualization and interpretation.
Accepted specimen types include fresh and formalin
fixed tissues, buccal swabs and fibroblasts. Of note, the
somatic analysis of a peripheral blood sample was
inappropriate in most cases.
Results: Clinical NGS analysis was performed on
specimens from 306 patients. The age of patients ranges
from newborn to 62 years, with an average age of 7.4
years. About half of the patients were under the age of
four and 18% were tested within their first year of life. A
total of 180/306 patients had reportable genetic variants,
85/180 had non-PIK3CA variants. Among those, 71/85
had a pathogenic (P) or likely pathogenic (LP) variant,
and 14/85 had variant of uncertain significance (VUS). A
total of 92 non-PIK3CA variants were identified in 23
genes in those 85 cases: 73/92 were P/LP and 19/92
were VUS. Half of the 92 variants were identified in six
genes: GNA11 (12/92), KRAS (8/92), PIK3R2 (7/92),
GNAS (7/92), GNAQ (6/92), and MTOR (6/92). Sanger
sequencing was performed on a secondary non-affected
specimen (peripheral blood) on 62/85 cases as a means
to confirm the origin of the causative variant. Among
those, 39/62 (63%) cases had negative results, indicating
the somatic origin of the variants, 23/62 (37%) cases had
positive results, indicating the germline or multi-tissue
mosaic origin. In correlation, 63% (37/59) variants are
with variant allele frequency (VAF) < 0.2 and 37% (22/59)
with VAF>0.2.
Conclusion: Our cumulative clinical experience
demonstrates the utility of NGS testing for segmental
overgrowth and related syndromes. With new clinical
trials for treatment that are often gene or variant specific,
somatic testing for causative variants in patients with
overgrowth syndromes has become an important part of
clinical care.
OR06. Validation of Immunohistochemistry by a
Monoclonal Antibody CAL2 on Trephine Biopsy for
CALR Mutation Detection in Myeloproliferative
M.K. Ng1, C.W. Lam1, C.H. Au1, B.K. Ip1, M.Y.R. To1, G.
Choy1, H.Y. Lam1, T.L. Chan1, S.K.E. Ma1
1Hong Kong Sanatorium & Hospital, Department of
Pathology, Hong Kong, Hong Kong
Introduction: Calreticulin (CALR) mutation is a major
diver genetic event in essential thrombocythemia (ET)
and primary myelofibrosis (PMF). Mutations in CALR
lead to 1+ base pair frameshifts and generate a novel
amino acid C-terminus. The monoclonal antibody CAL2
binds to the novel epitope of CALR mutant protein and
allows the recognition of CALR mutations. The objectives
of this study were to (1) validate the IHC detection of
CALR mutation detection by CAL2 in Myeloproliferative
Neoplasm (MPN) cases with known mutation status; (2)
to develop an IHC scoring system for CALR mutation
detection in the MPNs.
Methods: A total of 70 bone marrow (BM) trephines were
selected for paraffin CAL2 IHC, including 31 known
CALR mutations (22 ET and 9 PMF) and 31 known
JAK2V617F mutations (22 ET and 9 PMF). Molecular
testing for CALR mutation was performed by PCR
Sanger sequencing. JAK2V617F mutation was confirmed
by allele-specific PCR and PCR followed by restriction
enzyme digestion analysis (PCR-RFLP). Individual cases
were confirmed by fragment analysis or next-generation
sequencing as indicated. A known CALR mutated
specimen was used as an internal positive control, while
2 normal BMs and 6 cases of other myeloid neoplasms
and 2 CML-CP) were also used as negative control.
Mutant CALR mouse monoclonal antibody (CAL2,
Dianova, Germany) was titered at 1/100 by following
manufacturer's instructions. Slides were stained on
BOND III automated immunostainer (25 mins epitope
retrieval, 30 mins antibody incubation, and 8 mins
OptiView detection plus amplification). We developed a
scoring system calculated by megakaryocytes (MK)
signal strength (0-3+) times positive ratio of index (out of
1.0) for calling CALR mutation.
Results: Among 31 known CALR mutations cases with 9
various types of CALR mutations including 52 bp deletion
and 5 bp insertion, 22 positives with strong CAL2
cytoplasmic staining (intensity 2-3/3) and 8 positives with
weak CAL2 cytoplasmic staining (intensity 1) were
recognizable in the MK. The mean CAL2 IHC score was
1.87, 95% CI [1.56, 2.18]. There were no significant
differences between ET and PMF. 1/31 cases with
multiple clones of 2 CALR mutations in cis was negative
by CAL2 IHC. CAL2 staining was negative in 31 known
JAK2V617F cases, as well as in 2 normal BMs and 6
cases of other myeloid disorders. The CAL2 IHC results
correlated well with the molecular status.
Conclusions: The specificity and sensitivity of CAL2 IHC
were 100% and 97% respectively. CAL2 IHC is an
accurate method to identify CALR mutations in MPN.
While CAL2 antibody can recognize CALR mutations with
1+ base pair frameshift, a limitation is failure to detect
CALR mutation that results in non-canonical amino acid
sequence, albeit of uncertain pathogenicity. We have
successfully validated CAL2 IHC and the scoring scheme
is ready for clinical use on BM trephine. IHC by CAL2
antibody is a simple, quick and cheap method to identify
CALR mutation in MPN.
OR07. Comparison of a 640-Gene Next Generation
Sequencing Panel to a Virtual 41-Gene Panel for the
Diagnostic Utility in Unexplained Cytopenia
G. Zheng1, P. Chen1, L. Haley1, L. Jiang1, A.
Pallavajjalla1, M.-T. Lin1, C. Gocke1
1Johns Hopkins University, Baltimore, MD, United States
Introduction: The diagnostic utility of somatic mutations
for cytopenia is unclear: myeloid neoplasms including
myelodysplastic syndrome are clonal, but clonal
hematopoiesis can be found in healthy individuals. We
examined a large cohort of well-studied patients with
cytopenias with next generation sequencing to determine
S5 The Journal of Molecular Diagnostics
its diagnostic utility, and also assessed the impact of
panel size on diagnostic performance.
Methods: This cohort study comprised 207 patients with
cytopenias and a specimen (blood or marrow) submitted
to a CLIA-certified molecular pathology laboratory for
640-gene NGS from 09/2015 to 03/2018. Patients with
diagnoses of acute leukemia, lymphoma, or history of
cytotoxic therapy were excluded. Using a combination of
marrow morphology and cytogenetics as a diagnostic
“gold standard” for MDS, we assessed the diagnostic
utility of somatic mutations detected by the 640 gene
NGS panel and a virtual panel of 41 genes which are
commonly mutated in MDS and aplastic anemia.
Results: 586 somatic mutations were identified in 201
patients with definitive diagnosis: 102 of 105 MDS
patients, 20 of 34 aplastic anemia patients, and 27 of 52
patients with no evidence of a primary marrow disorder
showed somatic mutations. The sensitivity of the large
640-gene NGS panel for MDS was 98.3%. When the
variant allele frequency for one or more variants was set
at 20% the positive predictive value (PPV) for MDS was
94.9% with a specificity of 94.2%. Presence of two or
more somatic mutations at any variant allele frequency
(VAF) in a specimen showed a PPV of 76% for MDS,
which increased to 95.6% when the VAF cutoff was set to
10%. While the virtual 41 gene panel (Figure 1) showed a
mild decrease of sensitivity (96.5% versus 98.3%), a
better specificity was observed when VAF was set at
20% (98.8% versus 94.2%). Remarkably, the presence of
two or more somatic mutations at VAF 10% showed
100% specificity for MDS with the 41 gene panel.
Mutation analysis can also predict high versus low grade
MDS: mutations in TP53 and RUNX1 are associated with
high-grade MDS, while SF3B1 mutations are, as
expected, associated with MDS with ring sideroblasts. In
addition PIGA mutations were present exclusively in
aplastic anemia patients.
Conclusions: The 41 gene panel showed better
disgnostic specificty than the 640 gene panel. Both
panels showed high sensitivity in excluding MDS, and the
presence of a mutation with a VAF 20%, or two or
more mutations with a VAF10%, was highly predictive
of MDS. Presence of PIGA mutations was highly
predictive of aplastic anemia. Such mutation analysis has
the potential to improve the diagnostic approach and
accuracy for unexplained cytopenias.
[Figure 1. Somatic Mutations Identified by a virtual 41-
gene NGS panel in 201 cytopenia patients]
OR08. T-Cell Clonality Testing by Next Generation
Sequencing Facilitates the Initial Diagnosis and
Disease Monitoring of Cutaneous T-Cell Lymphomas
J. Yao1, M. Syed1, C. Ho1, W. Yu1, L. Maciag1, K.
Petrova-Drus1, C. Moung1, A. Zehir1, M. Roshal1, A.
Dogan1, K. Nafa1, M. Arcila1
1Memorial Sloan Kettering Cancer Center, New York, NY,
United States
Introduction: The diagnosis of cutaneous T-cell
lymphomas (CTCLs) is often challenging based on
morphology and immunohistochemistry. Clonality
assessment of T cell receptor (TCR) rearrangement is a
valuable adjunct in routine clinical work up; however,
current gold standard methods by fragment analysis (FA)
have distinct limitations due to low sensitivity and
specificity. Next generation sequencing (NGS) offers
unique advantages over traditional assays as it provides
sensitive and sequence specific data. This allows more
accurate initial diagnosis of cutaneous lesions as well as
highly sensitive assessment of disease involvement
beyond skin and the possibility to monitor disease post
treatment. Here we describe our clinical implementation
of TCR gamma gene (TRG) based NGS clonality testing
in the diagnosis and monitoring of CTCLs.
Methods: Clinical samples submitted for routine TRG
clonality testing were included in this study, including
formalin fixed paraffin embedded tissue, peripheral blood
and bone marrow. Testing was performed using the
LymphoTrack (LT) TRG kits (Invivoscribe) followed by
sequencing on Illumina MiSeq instrument. Data was
analyzed by LymphoTrack MiSeq Software and in-house
developed MSK-LymphoClone bioinformatics pipelines.
Performance characteristics were evaluated by directly
comparing to FA assay (BIOMED-2 primers) and flow
cytometry (FC).
Results: A total of 47 clinical samples from 33 patients
were analyzed, including Sézary syndrome/mycosis
fungoides (22), cutaneus gamma-delta T cell lymphoma
(3), anaplastic large cell lymphoma/lymphomatoid
papulosis (2), CD4+ small/medium T-cell
lymphoproliferative disorder (2), CD8+ aggressive
epidermotropic T-cell lymphoma (1), primary cutaneus
perihernial T-cell lymphoma (1), and CTCL, unclassifiable
(2). Among 34 initial characterization samples, 11 (32%)
were diagnosed as CTCLs after morphologic and
immunophenotypic assessment and a clone was
confirmed by both FA and LT. In 23 cases (68%), a
diagnosis of atypical lymphohistiocytic infiltrate was
rendered. TRG clonal rearrangement was detected in
16/23 (70%) by FA and 22/23 (96%) by LT. Among nine
monitoring samples with corresponding FC results, 9/9
and 6/9 were positive by LT and FC, respectively.
Conclusions: TRG NGS clonality testing demonstrated
higher clonal detection rate compared to FA and FC.
Improved clonal detection at initial assessment facilitated
the diagnosis of CTCLs in 26% more cases compared to
FA. The unique ability to search for diagnostic clonal
sequences enables more sensitive and specific staging
and monitoring in the subsequent specimens.
S6 The Journal of Molecular Diagnostics
OR09. A Noise-Cancelling Method Based on
Information Theory for ctDNA Variant Calling
H. Chen1, F. Zhang1, M. Mao1
1Seekin Inc., Shenzhen, China
Introduction: Circulating tumor DNA (ctDNA) is
fragmented cell-free DNA (cfDNA) in plasma, originating
from tumor cells rather than normal somatic cells. Since it
carries great potential for noninvasive early diagnosis,
progression monitoring, and prognosis, ctDNA soon
arose as one of the hottest topics in cancer research now
that NextGen Sequencing makes it feasible to identify
ctDNA in clinically acceptable cost and time. However,
due to low content of ctDNA in total cfDNA, a challenge
we confronted in clinical application was how to identify
meaningful low-frequency somatic variants carried by
ctDNA from chaotic noise introduced by sample
preparation and sequencing.
Information theory has been proven powerful to handle
signal-noise problems in various disciplines. Particularly
in bioinformatics studies, it has achieved competitive
accuracies in challenging signal process problems such
as identifying differential gene expression and aligning
distantly related protein sequences. Here we implement a
method based on knowledge of information theory to
cancel the noises in ctDNA sequencing data and amplify
the true variant signal.
Methods: Data preparation: SeqMaker
( was used to
generate simulated ctDNA sequencing data containing
73 single nucleotide polymorphism (SNP) variants and
their brinks of 1000bp in both upstream and downstream
with average depth of 1000x. The frequencies of these
variants varied from 0.001-0.1 and the sequencing error
rate was 0.001. Simulated data were then trimmed by
fastp and were then aligned by BWA (0.7.17-r1194-dirty).
The output was converted by Samtools into BAM and
PILEUP files.
Variant calling and performance benchmarking: The
information theory-based algorithm gave each base
position a score derived from the base frequencies in the
output PILEUP files. The detail of this score will be
provided in following publications. Then all base positions
were ordered by their score. Applying stepwise cutoffs on
these ordered positions, the true positive rate and the
false positive rate were determined. The receive operator
characteristic (ROC) figure summarizes all results.
Results: The ROC curve of our algorithm is displayed in
Figure 1. In this simulated ctDNA data, our algorithm can
achieve the true positive rate of 95% with the false
positive rate of 0.2% when calling variants. The Area
Under Curve (AUC) is 0.986.
Conclusions: Our information theory-based method can
cancel sequencing noises and accurately identify low-
frequency true variants in ctDNA samples, which could
improve diagnosis accuracy in clinical practice.
[Figure 1. ROC curve of our information theory-based
algorithm on simulated ctDNA data]
OR10. Clinical Implications of Cytogenetic
Heterogeneity in BCR-ABL1 Fusion Positive Adult B
Cell Acute Lymphoblastic Leukemia
X. Lu1, L.J. Medeiros,2
1Northwestern University Feinberg School of Medicine,
Pathology, Chicago, IL, United States, 2The University of
Texas MD Anderson Cancer Center, Hematopathology,
Houston, TX, United States
Introduction: Cytogenetic heterogeneity is frequently
observed in adults with t(9;22)(q34.1;q11.2)/BCR-ABL1
positive B cell acute lymphoblastic leukemia (Ph+B-ALL).
We retrospectively investigated the cytogenetic
heterogeneity and correlated additional chromosomal
abnormalities (ACAs) with clinical outcomes in 144 adults
with Ph+B-ALL.
Methods: There were 85 (59.1%) men and 59 (40.9%)
women with a median age of 54-years. Patients were
sorted into 6 subgroups based on the karyotypic findings
in stemline including: 32 (22.2%) with t(9;22) as the sole
aberration, 23 (15.9%) with t(9;22) plus one ACA, 26
(18.1%) with t(9;22) as part of a complex karyotype, 18
(12.5%) showed a variant t(9;22), 30 (20.8%) showed
t(9;22) as the stemline with ACAs in the sideline(s), and
15 (10.4%) had the t(9;22) and hyperdiploidy. Of these,
89 cases had one, 41 had two and 14 had three
clones, respectively.
Results: The median follow-up was 25.8 months. The
median overall survival (OS) was 25.6 months and the
median relapse-free-survival (RFS) was 20.6 months.
Patients with variant t(9;22) had a poorer OS and RFS
when compared with all other subgroups combined (P =
0.0018 and 0.0049, respectively). In addition, patients
with two or more clones had poorer outcomes than those
with one clone (P = 0.0427). Multivariate analysis
confirmed that variant t(9;22) and clone numbers are the
independent negative risk factors.
Conclusion: These data suggest that traditional
karyotypic analysis of adult Ph+B-ALL identifying
cytogenetic heterogeneity is helpful for the risk
stratification of these patients.
S7 The Journal of Molecular Diagnostics
Infectious Diseases
OR11. Incidental Detection of Malignancies in
Metagenomics Cell-Free DNA Testing
W. Gu1, E. Talevich2,3, A. Gopez1, S. Federman1, Y.
Sucu1, S. Arevalo1, M. Gottschall1, Z. Qi1, L. Wang1, B.
Karlon1, T. Kurtz1, J. Yu1, I. Yeh2, J. DeRisi4,5, S. Miller1,
C. Chiu1, J. Simko2
1University of California, San Francisco, Laboratory
Medicine, Menlo Park, CA, United States, 2University of
California, San Francisco, Pathology, San Francisco, CA,
United States, 3DNANexus, San Francisco, CA, United
States, 4Chan Zuckerberg Biohub, San Francisco, CA,
United States, 5UCSF, Biochemistry, San Francisco, CA,
United States
Introduction: Metagenomic next-generation sequencing
(mNGS) is an unbiased test for detecting DNA pathogens
through whole genome sequencing of cell-free DNA
(cfDNA) in plasma and body fluids. At University of
California San Francisco (UCSF) Clinical Laboratories,
we launched a clinically validated mNGS test for
undiagnosed meningitis and encephalitis using
cerebrospinal fluid and are validating a plasma mNGS
test for sepsis using cfDNA. In addition, we apply cell-
free metagenomics as a general approach to body fluids
suspected of infection. Using discarded human host data
that account for >99% of the sequencing reads, we found
genome-wide copy number variation signatures caused
by tumor derived cfDNA.
Methods: Clinical samples consisting of patient body
fluids and plasma were identified due to clinical suspicion
or laboratory confirmation of infection or malignancy. In
both blood and body fluids, we reduced the majority of
leukocyte human by centrifugation and sequenced only
the supernatant or plasma. We then created sequencing
libraries and performed whole genome sequencing. A
similar protocol was used for formalin-fixed, paraffin-
embedded (FFPE) material obtained from primary tumor
tissue. The fastq output was aligned to hg38 and binned
into segments of the human genome for read counting
and normalization.
Results: In the initial set of plasma samples from
patients newly admitted for suspected sepsis, 4/9 plasma
from metastatic cancer patients showed at least five large
(>20 Mb) copy number gains or losses. The positive
cases were validated based either on tissue copy number
variant (CNV) detection or cytogenetics for hematological
malignancies. All body fluid samples (e.g. pleural fluid,
peritoneal fluid, spinal fluid) that were confirmed positive
for malignancy on cytology also carried at least 5 large
(>20 Mb) copy number gains or losses. In addition, we
found five examples of positive tumor signatures in body
fluid specimens that had resulted 'benign' on cytology.
For a subset of the samples that had correlated tissue
blocks of the tumor available or had cytogenetic data, the
same CNV pattern was consistently found in a primary
Conclusions: We show that incidental data from cfDNA
metagenomic testing can pick up large CNVs in cfDNA.
This is particularly useful in cases where an undiagnosed
infectious and/or malignant etiology overlaps as
possibilities on a clinical differential. The high specificity
of requiring multiple large CNVs at a high tumor fraction
(>5%) will mitigate the risk of false positives. This test
may also be useful for monitoring the vast majority of
tumors in different body sites, working especially well for
metastatic disease with high tumor burdens. Larger
studies are needed to validate this finding across a
broader range of sample types and tumors and to assess
for clinical utility.
OR12. Comparative Whole-Genome Analysis
between Pregnant Women-Colonizing and Neonate-
Infecting Group B Streptococcus Isolates
L. Wu1, F. Wang1, R.E. Nunez2, W. Qiu2, Y. Tang3, Y.
1Bao an Maternal and Child Health Hospital, Jinan
University, Shenzhen, China, 2City University of New
York, Hunter College, New York, NY, United States,
3Memorial Sloan Kettering Cancer Center, New York, NY,
United States
Introduction: Group B Streptococcus (GBS) is a
frequent resident of the vaginal tract in pregnant women
and a major cause of invasive infection in neonates. GBS
is readily transmitted from mother to neonate; however,
molecular mechanisms for the transmission and invasion
from mothers to their neonates remain unknown.
Methods: In this study, 48 GBS isolates were collected
for whole-genome sequencing by using the Illumina next-
generation sequencing technologies and Five GBS
genomes have been fully assembled by PacBio,
Including 26 invasive (high-virulence, HV) isolates from
bloodstream-infected neonates and/or vaginal/mother
milk from colonization mothers, and 22 colonizing (low-
virulence, LV) isolates recovered from vaginal
colonization mothers without invasive infection in
We applied Align short reads to reference genome
(B111) by bwa, and call variants by samtools mpileup;
and annotate and extract single nucleotide
polymorphisms (SNPs) by customized BioPerl scripts.
The genome sequences between the mothers and
neonates in the four pairs were virtually identical. We
further compared the genomes between the HV and LV
isolates recovered from mothers with and without
subsequent neonate infections.
Results: In total, 34,395 variants, and 22,893 SNPs were
analyzed, all 48 GBS isolates genomes divided into five
major clades, Clade 1 consists of a significantly large
number of HV isolates (19 HV isolates included, p=6.7e-4
by Fisher's test) and other clades consist of more
balanced mixed isolates. Opportunistic infections occur in
all clades (no fixed diffs between invasive and colonizing
isolates) and no single gene or SNP was found to
correspond to all HV or LV (no fixed diffs between
invasive and colonizing isolates). Recombination (more
than mutation) has caused increase of virulence in the
HV clade (we used LDhat to run recombination hotspot
There are 2-3 loci that stand out as the most promising
virulence factor (a large number of nonsynonymous
S8 The Journal of Molecular Diagnostics
changes that are statistically significant), including:
LPXTG-domain surface protein (“CDH81_02355"), which
is a known virulence factor in Gram+ bacteria, and
phosphoribosylformylglycinamidine synthase
(“CDH81_07310"). This suggests that these are HV
factors that could be used as diagnostic markers and that
this is deserving of further research.
Conclusion: There are a group of GBS clones
circulating exclusively in China. Some clades are
associated with high-virulence in neonatal patients.
Multiple SNPs and genes are associated with high-
virulence, in particular the adhesion known as pullulates.
OR13. Suppressor of Cytokine Signaling (SOCS)-3
Downregulation Is Associated with Increased
Proinflammatory Responses in Diabetic Individuals
with M. tuberculosis Infection
K. Iqbal1, M. Rottenberg2, Z. Hasan1, M. Irfan1, Q.
1The Aga Khan University, Pathology and Laboratory
Medicine, Karachi, Pakistan, 2Karolinska Institute,
Stockholm, Sweden
Introduction: Pakistan ranks 5th in high tuberculosis (TB)
burden countries and 6th in countries with high burden of
diabetes mellitus (DM). Individuals with diabetes have
been shown to exhibit dysregulation of host protective
cytokines such as, IFN-γ, TNFα and IL6. Suppressor of
cytokine signaling (SOCS) molecule 3 plays a role in
maintaining balance between pro and anti-inflammatory
cytokines in patients with TB. Given the co-incidence of
both diseases we investigated the role of SOCS3 in
modulation of host immunity in individuals infected with
the pathogen, Mycobacterium tuberculosis (MTB),
comparing latent (LTB) and active disease.
Methods: Peripheral blood mononuclear cells (PBMCs)
were isolated from healthy controls (EC, n=20), those
with LTB (n=10), DM (n=13), DM with LTB (n=15) and TB
patients (n=15). PBMCs were stimulated with PPD
(10µg/ml) for 18 hours. Cell supernatants were collected
and tested for Th1/Th2 Cytokines. RNA was extracted
from stimulated PBMCs and mRNA used for RT-PCR
based analysis of IFNγ, TNFα, IL6 and SOCS3 genes.
Results: Measurement of stimulated cell supernatants
showed increased levels of IFN-γ in DM-LTB as
compared with EC (p=0.01), LTB (p=0.03), DM (p=0.01)
and TB (p=0.03) cases. IL12 and TNFα levels were
increased in DM-LTB as compared with LTB (p value:
IL12, 0.005; TNFα, 0.006), DM (IL12, p=0.021; TNFα,
0.019). Levels of IL6 were increased in DM-LTB as
compared with EC (p=0.001) and LTB (0.021).
Coordinately, we observed increased IFN-γ (p=0.033)
and IL6 (p=0.01) mRNA expression in patients with DM-
LTB as compared with EC; also, increased TNFα mRNA
expression as compared with LTB (p=0.034) cases.
SOCS3 mRNA expression was decreased in DM-LTB as
compared with LTB. In patients with TB, IFN-γ mRNA
levels were increased as compared with EC (p=0.0457),
while SOCS3 mRNA levels were decreased as compared
with LTB (p=0.0188).
Conclusion: We observed increased pro-inflammatory
response in diabetics with latent TB to be associated with
a downregulation in SOCS3 expression. As SOCS3 is
required for mycobacterial clearance, our data suggests
that this could be a mechanism which increases
susceptibility to active TB in DM cases who are infected
with MTB.
OR14. High Prevalence of NDM-1-Producing
Enterobacter cloacae from Three Tertiary Hospitals
in China
B. Huang1, Y. Cai1, C. Chen2, M. Zhao3, X. Yu1, K. Liao1,
K. Lan2, P. Guo1, W. Zhang2, X. Ma1, Y. He1, J. Zeng2, L.
Chen4, W. Jia3, Y. Tang5
1Department of Laboratory Medicine, The First Affiliated
Hospital of Sun Yat-sen University, Grangzhou, China,
2Department of Laboratory Medicine, The Second
Affiliated Hospital of Guangzhou University of Chinese
Medicine, Grangzhou, China, 3Department of Laboratory
Medicine, Ningxia Hospital of Ningxia Medical University,
Yinchuan, China, 4Public Health Research Institute
Tuberculosis Center, New Jersey Medical School,
Rutgers University, Newark, NJ, United States,
5Department of Laboratory Medicine, Memorial Sloan
Kettering Cancer Center, New York, NY, United States
Introduction: Enterobacter cloacae has recently
emerged as one of the most common carbapenem-
resistant Enterobacteriaceae. The emergence and
spread of NDM-1-producing E. cloacae have posed an
immediate threat globally. Here, we investigated the
molecular characteristics of carbapenem-resistant
Enterobacter cloacae (CREL) isolates collected from
three tertiary hospitals in certain regions of China from
2012 to 2016.
Methods: Eighty-six nonduplicate CREL isolates were
collected from Southern (e.g. Grangdong) and
Northwestern (e.g. Ningxia) China. Species identification
and antimicrobial susceptibility testing were performed
using a VITEK-2 system. Susceptibility to carbapenems
were also confirmed using broth microdilution method.
Isolates were screened for antibiotic resistant genes by
PCR, and expression of ompC and ompF was
determined by qRT-PCR. Genetic relatedness was
performed by pulsed-field gel electrophoresis (PFGE),
and selected isolates were subjected to whole-genome
sequencing. Multilocus sequence typing (MLST) and
plasmid-typing were also mined in our study.
Results: Among the 86 CREL isolates, 50 (58.1%) were
detected as carbapenemase producers. NDM-1 was the
dominant carbapenemase (46.5%), followed by IMP-26
(4.7%) and IMP-4 (3.5%). Notably, we identified the first
NDM-1 and IMP-1 co-producing E. cloacae, carrying
plasmids of three incompatibility (Inc) groups, including
IncHI2, IncHI2A, and IncN. In addition, Most strains
showed decreased expression of ompC and/or ompF,
and contained a broad distribution of extended-spectrum
β-lactamases (ESBLs) and AmpC β-lactamase genes.
The high-throughput sequencing demonstrated that
seven blaNDM-1-harboring plasmids contained a highly
conserved gene environment around blaNDM-1 (blaNDM-1-
bleMBL-trpF-dsbD-cutA1-groES-groEL). Meanwhile, ltrA
and QacEΔ1 genes were frequently observed
S9 The Journal of Molecular Diagnostics
downstream blaIMP-26 and blaIMP-4. All the CREL isolates
could be divided into 60 PFGE clusters. Nineteen blaNDM-
1-positive E. cloacae isolates obtained from Ningxia had
the same PFGE pattern (Type 1) and were classified to
ST78 in clonal complex 74 (CC74) by MLST analysis.
The replicon typing indicated that IncX3 plasmids
mediated the dissemination of blaNDM-1 within these
homologous strains.
Conclusions: These findings suggested that different
molecular mechanisms, including carbapenemase,
ESBLs and AmpC β-lactamases plus the loss of porins,
have contributed to carbapenems resistance in these
CREL isolates. We first reported an outbreak of NDM-1-
producing E. cloacae ST78 with contribution of IncX3
plasmids in northwestern China. There´s an immediate
need to to heightened surveillance efforts attentively to
prevent and control the further spread of NDM-1 in
China. (Y.C., C.C., MZ, and X.Y. contributed equally to
this article)
OR15. Application of a Multiplex PCR to Detect
Intestinal Parasite Infections in Antenatal Anemia
V. Gupta1,2, P. Deshpande1,2, Q. Zahiruddin3, A.
1Datta Meghe Institute of Medical Sciences, Molecular
Biology and Epidemiology Laboratory, Wardha, India,
2Datta Meghe Institute of Medical Sciences, Pathology,
Wardha, India, 3Datta Meghe Institute of Medical
Sciences, Community Medicine, Wardah, India, 4Datta
Meghe Institute of Medical Sciences, Community
Medicine, Wardha, India
Introduction: Anemia and poor nutrition status,
especially in pregnant women, are partially attributable to
intestinal parasitic infections. The majority of the global
burden is considered due to the five main parasites:
Ascaris lumbricoides, Necator americanus, Ancylostoma
dudonale, Trichuris trichiura, Giardia duodenalis, and
Entamoeba histolytica. Diagnosis of these parasites has
always relied on the classical microscopic examination of
stool samples as it is relatively simple to perform and
does not require expensive laboratory equipment.
However, stool microscopy is highly observer dependent
and has several shortcomings that greatly affect the
efficacy of current management of intestinal parasitic
infections in pregnant women. The study aim to assess
the utility of multiplex PCR for detection of intestinal
parasites as compared to microscopy in establishing
intestinal parasitic infections in antenatal anemic.
Methods: Thirty-four antenatal anemic subjects, enrolled
from neighboring villages of Wardha district of India as
part of pilot study funded by Public Health Foundation of
India in collaboration with Department of Science and
Technology, India.
Samples were collected and stored in 5% potassium
dichromate. DNA extraction was done using conventional
protocol. Extracted DNA was subjected to multiplex real
time PCR using TaqMan probes, forward, and reverse
primers for the following species Ascaris lumbricoides,
Ancylostoma dudonale, Trichuris trichiura, Necator
americanus, Giardia duodenalis, and Entamoeba
histolytica was designed.
[Primers and Probes for multiplex real time PCR used to
detect different parasites]
PCR was performed using PikoRealTM Real time PCR
System by Thermo Fisher Scientific. Final reaction
volume was set at 20µl, with UNG incubation at 50ºC for
2 min, polymerase activation at 95ºC for 10 min, followed
by 45 cycles of denaturation (95ºC for 15 sec) and
annealing and extension (60ºC for 1 min). The results
were compared with standard microscopy.
Results: The age range of antenatal anemic patients
was 18-30 years median age 24 years. 16/34 (41%)
patients were positive for one or the other parasite.
Maximum number 11/34 (32%) had infection with Ascaris
lumbricoides, 7/34 (12%) were positive for Ancylostoma
dudonale, 1/34 was positive for Entamoeba histolytica.
4/34 had multiple infection with Ascaris lumbricoides and
Ancylostoma dudonale. On microscopy only two cases
(Ascaris lumbricoides and Entamoeba histolytica each)
were positive and none was positive for co-infection.
Conclusion: The real-time PCR-based diagnosis
outperforms microscopy and the multiplex PCR approach
was also superior in terms of detection with multiple
infections and can be used for detection and
management in antenatal anemia.
S10 The Journal of Molecular Diagnostics
OR16. Identifying True Somatic Variants in Cancer
Using Machine Learning
C. Wu1, X. Zhao1, M. Welsh1, K. Costello2, K. Cao1, M.M.
Li1,3,4, M. Sarmady1,3
1Children's Hospital of Philadelphia, Division of Genomic
Diagnostics, Philadelphia, PA, United States, 2Temple
University, College of Science and Technology,
Philadelphia, PA, United States, 3Perelman School of
Medicine, University of Pennsylvania, Department of
Pathology and Laboratory Medicine, Philadelphia, PA,
United States, 4Perelman School of Medicine, University
of Pennsylvania, Department of Pediatrics, Philadelphia,
PA, United States
Introduction: Molecular profiling has become
increasingly essential for tumor risk stratification and
treatment selection. However, the diversity of somatic
events, technical artifacts and genomic complexity make
the differentiation of reportable variants from artifacts a
challenge. Variant refinement via manual inspection is
costly for clinical labs and is subjective to personal bias.
Here, we present a machine learning-based classifier
developed at Children's Hospital of Philadelphia (CHOP)
to distinguish artifacts from bona fide Single Nucleotide
Variants (SNV)s of tumor samples.
Methods: All of the SNVs used in this study were from
CHOP cancer panels and were manually reviewed and
labeled as either true positive or artifact. An array of
sequencing attributes were calculated for each SNVs,
such as coverage, strand bias and mapping quality.
Feature selection was performed to remove attributes
that provided little separation between the two classes
such as mapping quality. A random forest classifier was
trained on a training set of 3367 SNVs from 84 tumor
samples. The classifier achieved 100%
sensitivity/specificity and 0.98 F-1 score in a 10-fold
cross validation. Following this, an independent test set
of 2362 SNVs from 66 samples were used to fine-tune
the baseline model. During this process, a confidence
interval (CI) in the range of [0, 1] was assigned to each of
the variants by the model reflecting the certainty of the
classification. The range indicates confidence level of the
model for a variant being 100% an artifact 0.0 or real 1.0.
To achieve optimal performance, different weights were
assigned to the two classes to account for different
misclassification costs, i.e. a true variant being classified
as an artifact (false negative) could lead to a worse
adverse clinical consequence than a false positive
Results: We identified a model that yielded the optimal
performance on the test set. Additionally, a range of CI
scores was derived to label SNVs with CI scores within
[0.05, 0.9] “uncertain” to meet clinical assurance, which
would require manual inspection. The combination of the
classifier and the CI thresholds, achieved 100%
sensitivity/specificity with 11 SNVs labeled as “uncertain”
and zero misclassifications on the test set. We then
benchmarked its performance on an independent
validation set of 5634 SNVs from 141 samples. As a
result, our model achieved 100% sensitivity/specificity
with only 3% SNVs labeled as “uncertain” and zero
Conclusions: We presented a computational classifier to
identify true positive SNVs from tumor sequencing. The
classifier demonstrated high sensitivity/specificity, and
utility on a wide range of tumor samples. Overall, 90% of
the SNVs detected will receive a definite label and thus
be exempt from manual review. Implementing the model
can greatly reduce the hands-on time and hence improve
the efficiency without compromising the quality of the
clinical tests.
OR17. Development and Clinical Validation of a
Bioinformatics Pipeline for CNV Detection on Cancer
NGS Panels
K. Cao1, F. Lin1, F. Chang1,2, M. Welsh1, M. Li1,2, M.
1Children's Hospital of Philadelphia, Philadelphia, PA,
United States, 2Perelman School of Medicine, University
of Pennsylvania, Philadelphia, PA, United States
Introduction: Copy number variation (CNV) is the most
common genomic alteration in pediatric cancer that
directly impact patient care. NGS has emerged as the
method of choice to analyze a wide range of genes
known to be associated with various cancers. We have
launched multiple NGS panels targeting childhood
cancers including 13 hereditary cancer panels (128
genes), 9 bone marrow failure panels (159 genes), one
hematological cancer panel (118 genes), and one solid
tumor panel (238 genes). To optimize the detection of
CNVs based on the sequencing data, we developed an
accurate and efficient bioinformatics pipeline to dentify
and visualize CNVs allowing simultaneous detection of
SNVs, indels and CNVs.
Methods: To overcome the limitation of targeted panel
sequencing that lacks continuous coverage required for
the detections of CNV and to meet the needs of
identifying ploidy changes in cancers, thousands of single
nucleotide polymorphism (SNP) probes relatively evenly
covering the entire genomic regions of targeted genes
were added to each panel as the “backbones”. FastQ
files generated by HiSeq sequencer are first aligned to
GRCh37 assembly. CNVs are called by VarScan2 on the
data from patients and controls sequenced using the
same set of baits. B-allele frequency was generated
based on SNP calls from the backbone and the regions
of interest of the panel for each patient and compared to
that of the controls. CNVs were annotated with gene
symbols and chromosome locations. CNVs and B-alleles
were then post-processed and visualized by home-brew
software using R Shiny. The pipeline were designed for
analyzing multiple samples in parallel to increase analytic
efficiency and accuracy.
Results: To assess the performance of the pipeline, we
validated it on 30 different clinical samples with known
CNVs. The results showed 100% concordance to that of
our previously validated commercial software yet with
only 20% of the processing time. Analyzing multiple
cases in paralell also improves the accuracy of the CNV
calls by eliminating batch effect. In addition to detecting
S11 The Journal of Molecular Diagnostics
simple CNVs, the pipeline successfully detected complex
genomes with complicated del/dup/amp and ploidy
changes in neuroblastoma, rhabdomyosarcoma, and
osteosarcoma exemplified by the identification of a
pseudo-hyperdiploid genome in a near-haploid ALL
patient which would be indistinguishable from true
hyperdiploid ALL by cytogenetics or histology/flow
cytometry (figure 1).
Conclusions: We have developed a clinical
bioinformatics pipeline for CNV calling, annotation and
visualization for targeted NGS panels. The validation
results demonstrate 100% accuracy in identifying cancer
associated CNVs with increased efficiency and accuracy.
[figure 1 for caok g-amp 2019 abstract]
OR18. Pan-Cancer Repository of Validated Natural
and Cryptic mRNA Splicing Mutations
P. Rogan1,2, E. Mucaki1, B. Shirley2
1University of Western Ontario, Biochemistry, London,
ON, Canada, 2Cytognomix, London, ON, Canada
Introduction: Next generation sequencing continues to
reveal large numbers of novel variants whose impact
cannot be interpreted from curated variant databases or
through reviews of peer-reviewed biomedical literature.
This has created a largely, unmet need for unequivocal
sources of information regarding the molecular
phenotypes and potential pathology of variants of
unknown significance (VUS); in cancer genomes, such
sources are critically needed to assist in distinguishing
driver mutations from overwhelming numbers of
bystander mutations.VUS classification criteria highlight
the limitations in genome interpretation due to ambiguous
variant interpretation.
Methods: We present a major public resource of mRNA
splicing mutations validated according to multiple lines of
evidence of abnormal gene expression. Likely mutations
present in all tumor types reported in the Cancer
Genome Atlas (TCGA) were identified based on the
comparative strengths of splice sites in tumor versus
normal genomes and then validated by respectively
comparing counts of splice junction spanning and
abundance of transcript reads in RNA-Seq data from
matched tissues and tumors lacking these mutations.
Results: The comprehensive resource features 351,423
of these validated mutations, the majority of which
(69.1%) are not present in the Single Nucleotide
Polymorphism Database (dbSNP 150). There are
117,951 unique mutations which weaken or abolish
natural splice sites, and 244,415 mutations which
strengthen cryptic splice sites (10,943 affect both
simultaneously). 27,803 novel or rare flagged variants
(with < 1% population frequency in dbSNP) were
observed in multiple tumor tissue types. Single variants
or chromosome ranges can be queried using a Global
Alliance for Genomics and Health (GA4GH)-compliant,
web-based Beacon “Validated Splicing Mutations” either
separately or in aggregate alongside other beacons
through the public Beacon Network (http://www.beacon-, as well as
through our website
Conclusions: The Validated Splicing Mutation resource
should substantially contribute to reducing the number of
outstanding VUS in tumor (and possibly some germline)
genomes, and substantially increases the number of
functional variants with previously unappreciated
consequences to mRNA splicing, in particular, activation
of cryptic splice sites We previously (doi:
10.1038/srep07063) identified 988 TCGA breast cancer
mutations that significantly altered normal splicing (19%
of total). This database greatly expands the size of the
repository. Here, a higher ratio of rare or novel mutations
have been validated (24% of total were flagged). The
higher yield could be related to the same mutation being
present in multiple samples from the same tumor type
and other tumor tissues, which would be expected to
increase the probability of observing abnormally
expressed splice forms for the mutation.
OR19. Error Rate Normalization to Establish Position-
Specific Limits of Detection in Next Generation
Sequencing Assays
J. D'Angelo1, S. Ferguson2, Z. Xie1, L. Chamberlain1, A.
Carson1, J. Miller1
1Invivoscribe Technologies, San Diego, CA, United
States, 2LabPMM, San Diego, CA, United States
Introduction: Next-generation sequencing (NGS)
continues to be the method of choice for high throughput
genetic assays. Many assays employ a uniform assay-
specific limit of detection (LoD) that tends to be overly
conservative due to “worst-case” error and artifact rates
(generated by PCR, sequencing and alignment errors,
etc.). The process of differentiating sequencing artifacts
from real, low-frequency variants is extremely important
for clinical utility of NGS-based assays, especially as
clinicians push for tracking minimal residual disease
(MRD). The motivation for this study was to identify a
strategy for finding the inherent error-rate, at single base
resolution, for our MyMRD assay. This strategy improved
the specificity and sensitivity of our MyMRD assay and
can be implemented in any NGS-based assay.
Methods: Our study focused on Invivoscribe's MyMRD
assay panel and investigated the observed error rates at
all 53,308 targeted bases. We looked at each of these
positions independently to evaluate the observed
frequency of every possible single nucleotide variation
(SNV) and every observed insertion/deletion (indel) in
213 samples (36 samples from reference cell line
NA12878 and 177 contrived/clinical samples).
S12 The Journal of Molecular Diagnostics
We calculated various inter and intra-sample metrics
(e.g. minimums, maximums, average error frequencies)
independently for each SNV and indel to understand
overall assay trends as well as sample-specific trends at
single base resolution.
0% 0.30
0% 0.33
d/Clinical 0% 0.49
0% 0.10
[Table 1]
Table1: The analysis showed that error-rates widely vary
across different positions and across different variant
types (even within a single targeted base).
**Known variants (in NA12878) and high probability
germline variants and artifacts were filtered from these
analyses to focus on the background error-rate at
homozygous reference positions.
Conclusions: We were able to increase the sensitivity
and specificity of the MyMRD panel by creating non-
uniform LoDs that take into account variant specific error
rates at every targeted position. We improved the
sensitivity of the assay by decreasing the LoD (often
dropping from 0.5% to lower than 0.1%) and at >99% of
the targeted sites. However, we also identified potential
artifacts where the LoD was increased to avoid calling
false positives which would have arisen if a uniform LoD
was used. By increasing the LoD at these positions, we
were also able to improve the assay's specificity. Overall,
performing this analysis makes it easier to differentiate
sequencing artifacts from true low-frequency variants,
which is particularly important for MRD tracking
Solid Tumors
OR20. Clinical Implication of ctDNA Analysis in
Advanced Lung Carcinoma Patients Using Different
Technologies: Real-Time PCR and MALDI-TOF
P.-J. Lamy1, N. Lozano1, C. Becht2, L. Gambier1, N.
Pourel3, C. Alix-Panabieres4
1Inovie, Institut d'Analyse Génomique, Imagenome,
Montpellier, France, 2Centre de Cancérologie du Grand
Montpellier, Thoracic Oncology, Montpellier, France,
3Institut Sainte Catherine, Onco-Thoracic Unit, Avignon,
France, 4University Medical Centre of Montpellier,
Laboratory of Rare Human Circulating Cells, Montpellier,
Introduction: With liquid biopsy (LB) entering the routine
in personalized cancer treatment it is important to
understand the differences with regards to technical and
clinical sensitivity and mutation coverage inherent to
technologies for analyzing circulating tumor DNA (ctDNA)
and its implication for the patient care.
Methods: Circulating cell free DNA (cfDNA) extracted
from blood plasma (QiaAMP ctDNA, Qiagen) of 100 lung
carcinoma patients were assessed for DNA quality and
quantity using DNA 3K Assays on the LabChip GX Touch
24 (PerkinElmer). Each 10 ng cfDNA were analyzed for
somatic mutations detection in ctDNA using two
technologies: Real-time PCR test Cobas EGFR Mutation
Test v2 (Roche) that identifies 42 mutations in the EGFR
gene (63 patients tested), and UltraSEEK Lung Panel on
MALDI-TOF based MassARRAY (Agena Bioscience) that
covers 67 mutations in 5 oncogenes: EGFR (43), BRAF
(4), ERBB2 (2), KRAS (14) and PIK3CA (4) with a
sensitivity down to 0.1% (100 patients tested). All
samples were also analyzed with the UltraSEEK EGFR
Panel (Agena Bioscience) that surveys the 3 most
frequent activating mutations in EGFR and the resistance
markers EGFR T790M and EGFR C797S. A set of
controls was added to each run: 0.5% positive control
(EGFR cfDNA Multiplex), EGFR wild-type control
(Horizon Discovery) and no-template control.
Results: All cfDNA samples showed the expected
fragment-size distribution on LabChip and no significant
abundance of genomic DNA.
Four of the 63 (6%) samples tested on Cobas EGFR
Mutation Test v2 were invalid for analysis. Concordant
results were achieved for 85% of patients (50/59) for
overlapping markers in EGFR. With the UltraSEEK Lung
Panel 10 additional EGFR mutations (4*Exon 19 deletion,
3*T790M, 1*Exon 20 insertion, 1*L861Q/R, 1*G917A)
were detected in 8 of 59 patients (14%). 1 EGFR Exon
19 deletion was detected by Cobas only (2%). EGFR
T790M mutation was detected in 7 samples with Cobas
and in 10 samples with UltraSEEK (70% concordance).
UltraSEEK Lung Panel identified 7 non-EGFR mutations
(6*KRAS, BRAF) in 6 patients (10%). Across all 100
patient samples in EGFR 32*Exon 19 deletions,
18*T90M, 17*L858R, 5*G719X, 3*L861X and 1*S768I
could be identified with UltraSEEK Lung Panel as well as
9*KRAS and 2*BRAF V600E. The UltraSEEK EGFR
Panel confirmed all overlapping mutations in EGFR also
assessed with the UltraSEEK Lung Panel.
S13 The Journal of Molecular Diagnostics
Conclusion: This study showed that the UltraSEEK
Lung Panel improved sensitivity for ctDNA analysis in
comparison to Cobas EGFR Mutation Test v2, a method
widely used in clinical routine. The UltraSEEK Panel
could detect more EGFR primary or secondary
resistance mutations. With the use of a panel that covers
mutations in multiple genes, additional actionable
mutations to EGFR (KRAS, BRAF) could be also
detected in LB and used as a guidance for therapy.
These new data may lead to improve clinical sensitivity of
LB and its clinical utility in the management of lung
cancer patients.
OR21. Adjusting for Variation in Tumor Mutation
Burden Using a Cancer-Specific Threshold in Whole
Exome Sequencing
W. Song1, E. Fernadez1, K. Eng1, O. Elemento1, M.
Shah1, A. Sboner1
1Weill Cornell Medical College, New York, NY, United
Introduction: Tumor mutational burden (TMB) has
recently been shown to be an improved biomarker for
predicting response to immunotherapy, with high TMB
patients responding better to immunotherapy than low
TMB patients. Defining a patient as TMB high or TMB low
has not been standardized and often uses a pan-cancer
threshold. On the other hand, almost all previous
publications on TMB cut-off have used comprehensive
panel which covers a small portion of the whole
genome/exome, and no systematic whole exome
sequencing data has been analyzed to analyze the TMB
distribution among different cancer types. Hereby, we
aim to understand the differences between using a pan-
cancer threshold compared to a cancer-specific
threshold, as well as understand the clinical context of
Method: Using whole exome sequencing (WES) data
from primary tumors in The Cancer Genome Atlas
(TCGA) (n=3886) and advanced Weill Cornell Medicine
(WCM) samples (n=979), TMB status was determined by
using both a pan-cancer and cancer-specific threshold.
Survival curves and number of TMB high classifications
were used together to evaluate the differences between
the thresholds. Cox regression was performed to
understand other clinical variables that may influence the
use of TMB as a prognostic biomarker.
Result: The distribution of TMB varied between cancer
types. The cancer-specific threshold was able to adjust
for the different TMB distributions, while the pan-cancer
threshold was often too stringent. The dynamic nature of
the cancer-specific threshold resulted in more TMB high
classifications compared to the static pan-cancer
threshold. Additionally, the survival curves between the
cancer-specific and pan-cancer threshold were similar for
both TMB high and TMB low groups.
Conclusion: TMB is relative to the context of cancer
type, metastatic state, and disease stage. A cancer-
specific threshold results in more patients being classified
as TMB high while maintaining clinical outcome
comparable to pan-cancer threshold
[Distribution of TMB Among Different Cancer Types]
[Comparison of TMB-H Samples by WCM versus
Chalmers' Cut-off]
S14 The Journal of Molecular Diagnostics
OR22. Integrated Genomic Profiling of 289 Pediatric
Brain Tumors Uncovers Genomic Signatures
Significantly Impacting Patient Care
F. Lin1, L. Surrey1, A. Resnick2, P. Storm2,3, M. Luo1, D.
Gallo1, K. Cao1, X. Zhao1, S. MacFarland2, M. Santi1, A.
Waanders2, M. Li1,2
1Children's Hospital of Philadelphia, Department of
Pathology and Laboratory Medicine, Philadelphia, PA,
United States, 2Children's Hospital of Philadelphia,
Center for Childhood Cancer Research, Philadelphia, PA,
United States, 3Children's Hospital of Philadelphia,
Department of Pediatrics, Division of Neurosurgery,
Philadelphia, PA, United States
Introduction: Genomic mutation profiling plays a
significant role in brain tumor diagnosis, prognosis, and
treatment. Large Next-Generation Sequencing (NGS)
panels are rapidly replacing traditional single-gene or
focused multigene testing to provide more
comprehensive information about the tumor genome.
Here we describe genomic profiles of a large cohort of
pediatric brain tumors tested in a single clinical laboratory
using a custom-designed NGS panel, CHOP OncoMark
Solid, and their impact on patient care.
Methods: Brain tumor samples from patients 0-21 years
old over a 33-month period were subjected to NGS
testing. Genomic alterations were assessed using the
CHOP OncoMark Solid panel to interrogate 238 cancer
genes for single nucleotide variants (SNVs), indels, and
copy number variants (CNVs), and 110 fusion gene
partners for over 600 known and novel fusions. Identified
variants were categorized and reported according to the
AMP/ASCO/CAP guidelines. The clinical impact of the
results were evaluated based on their significance on
diagnosis, prognosis and treatment.
Results: 289 consecutive tumors were tested, the most
common of which were pilocytic astrocytoma (87/289),
medulloblastoma (24/289) and diffuse midline glioma
(21/289). Clinically significant genomic alterations were
identified in 93% of the tumors. 191 clinically significant
SNVs/indels in 73 genes were reported 323 times in 164
cases. The most frequent mutations were BRAF V600E
(32/323) in multiple tumor types and H3F3A K28M
exclusively in diffuse midline glioma. TP53 and NF1 were
second and fourth most mutated genes in this cohort.
More than 1500 CNVs were reported in 175 cases with
gain of 1q and 7q the most prevalent. 32 fusions were
detected in 97 cases with KIAA1549-BRAF the most
common. These results provided diagnostic evidence in
73.3% of patients including 9 patients whose diagnoses
were changed and 41 patients whose diagnoses were
further defined; prognostic evidence in 34.6% of patients,
and evidence for targeted therapy in 18% of patients.
Distinctive genomic profiles for specific tumor subtypes
were observed: 64% of pilocytic astrocytomas harbored
KIAA1549-BRAF, all diffuse midline gliomas were H3F3A
K28M positive with 52% of them showing H3F3A gain,
isochromosome 17q was highly enriched in non-
WNT/non-SHH medulloblastomas, and loss/cnLOH of
chr22 was observed in 100% of ATRT cases, 42% of
which displayed additional SMARCB1 mutations. 71
patients were suspected of carrying germline mutations,
of which 25 were evaluated and 11 confirmed. The
average tumor mutation burden (TMB) was 3.88
mutations/Mbp and all 5 cases with high TMB (>10
mutations/Mbp) harbored a germline mutation in MSH2
or MSH6.
Conclusion: Integrated genomic profiling of pediatric
brain tumors demonstrates significant impact on clinical
care in 78.2% of patients in our institution. Broad-base
genomic profiling should be implemented as part of the
routine patient care for pediatric brain tumors.
OR23. Biomarker Analysis on Limited Biopsy Lung
Cancer Samples in Tertiary Care Cancer Centre in
South India Basavatarakam Indo American Cancer
Hospital, Hyderabad, India
S. Sudha Murthy1, L. Nambaru2, S.J. Rajappa3, G.
Sandhya Devi1, S.A. Sahithi1, K.M. Mallavarapu3, F.
Daphne1, T.S. Rao4, C. Veeraiah5
1Basavatarakam Indo American Cancer Hospital,
Pathology & Laboratory Medicine, Hyderabad, India,
2Basavatarakam Indo American Cancer Hospital,
Laboratory Medicine, Hyderabad, India, 3Basavatarakam
Indo American Cancer Hospital, Medical Oncology,
Hyderabad, India, 4Basavatarakam Indo American
Cancer Hospital, Surgical Oncology, Hyderabad, India,
5Basavatarakam Indo American Cancer Hospital,
Radiology, Hyderabad, India
Introduction: Lung cancer diagnosis and management
has undergone a paradigm shift in the last decade with
multiple actionable mutations and genetic alterations
being detected. Biomarker studies in lung cancers from
India are limited owing to the different tissue handling
processes and lack of standardisation in the preanalytic
phase. We present the data of prevalence of EGFR
(Epidermal growth factor), ALKD5F3 (Anaplastic
lymphoma kinase; clone D5F3) and ROS1 in a cohort
from a tertiary care centre in south India seen over the
last two years.
Methods: EGFR testing was done by RT-PCR (Qiagen
Rotorgene) using Therascreen EGFR RGQ PCR kit,
ALKD5F3 testing was done by immunohistochemistry
(IHC) on Ventana Benchmark XT automated
immunohistochemistry platform and ROS1 testing by
Flouroscence in situ hybridisation (FISH) using Zytolight
break-apart probe.
Results: Of a total 515 lung biopsies performed. EGFR
mutation testing was done in 269 (52.2%) cases.
Morphology included 406 (78.8%) adenocarcinomas,
51(9.9%) adenosquamous carcinomas, 35 (6.7%)
squamous cell carcinomas and 23 (4.4%) of poorly
differentiated carcinomas. Overall EGFR mutation was
detected in 141(52.4%) patients. The predominant
mutations identified were Del 19 in 77 (28.6%) cases,
L858R in 52 (19.3%) cases and 102 (37.9%) cases were
of wild type. Dual mutations Del19+L858R were noted in
two cases, Del19+L861Q in one case, L858R+T790M in
one case and T790M+Del19 in two cases. The
biomarkers were mutually exclusive in majority of the
cases, however one case showed dual ALKD5F3
expression by IHC and EGFR mutation (insertion). One
S15 The Journal of Molecular Diagnostics
other case harboured Del19 mutation and ALKD5F3
positivity. 200 patients underwent ALKD5F3 testing by
IHC, of which 36 (18%) patients expressed ALK
positivity. 195 patients underwent ROS1 testing by FISH
of which five (2.56%) patients harboured ROS1 gene
T790M testing was done at the time of progression to test
for secondary resistance by RTPCR on biopsy samples
in 48 patients. Of these 26 patients (54%) harboured
T790M mutations indicating secondary resistance
pattern. Of these 26 patients, 17(65.3%) harboured Del
19 mutaion, six (23%) harboured L858R, two had a
report of mutation (done elsewhere) and one was wild
type (done elsewhere).
Conclusion: Triaging of tissue samples and coordination
between multidisciplinary teams is the key to successful
biomarker analysis and patient management in lung
OR24. Oncologists' Review of ctDNA EGFR Mutation
Testing Reports
Z. Deans1, G. Holland1, N. Iley1, F. Blackhall2, A.
Greystoke3, D. Talbot4, R. Butler5
1GenQA/UK NEQAS, Edinburgh, United Kingdom, 2The
Christie NHS Foundation Trust, Manchester, United
Kingdom, 3Newcastle upon Tyne Hospitals NHS
Foundation Trust, Newcastle upon Tyne, United
Kingdom, 4Oxford University Hospitals NHS Foundation
Trust, Oxford, United Kingdom, 5All Wales Genetic
Laboratory, Cardiff, United Kingdom
Introduction: GenQA is an External Quality Assessment
(EQA) Provider for Genomics which assesses laboratory
tests and interpretative content of reports. GenQA
provided an EQA for the testing of circulating tumour
DNA (ctDNA) for EGFR mutations in lung cancer
patients. The submitted reports were anonymously
reviewed by a panel of Consultant Oncologists to identify
the elements which were clear and unambiguous, and
those misleading and may cause patient harm if
misinterpreted. The findings were shared with the
laboratories to improve the reporting standard of liquid
biopsy test results.
Methods: Nine laboratories perform ctDNA testing for
EGFR mutations in lung cancer patients in the United
Kingdom, and all took part in this EQA and report review.
The EQA reports submitted for two clinical cases were
anonymised and sent to the reviewers to be rated for
ease of finding the result, understanding the test
methodology and clarity of the interpretation. Text which
the reviewer found comprehensive, and content which
was confusing or inaccurate were highlighted. Individual
laboratory reviews were provided to the laboratories.
Results: The review identified elements deemed to be
essential on the report; a clear statement of the mutation
testing result in a format which cannot be misinterpreted,
understandable test information and associated
limitations, the use of HGVS nomenclature cross
referenced to the common names e.g. T790M, a
condensed one page report and generic reference to
targeted treatment rather than naming individual drugs.
The reports submitted for a patient progressing on first
line EGFR TKI therapy with the primary EGFR mutation
and a resistance mutation detected in the plasma
sample, raised comments about reporting allelic
frequency and queried the clinical utility for this on the
report. The clinicians wished reassurance that ctDNA had
been tested as the primary EGFR mutation had been
The second report set was related to a patient relapsing
on EGFR TKI therapy with no EGFR mutations detected
in the plasma sample. The reviewers' responses were
focussed on the risk of a false negative result and the
ways the laboratories reported this varied, if reported at
all, and the recommendations on appropriate further
testing. The reviewers expressed a wish for a more
standardised approach to interpreting such results and to
emphasise the importance on highlighting that the
primary EGFR mutation had not been detected.
Conclusions: Laboratories report ctDNA EGFR mutation
testing in lung cancer patients using a multitude of
formats and terminology. The lack of standardisation
increases the risk of misinterpreting the result with the
potential to harm patients through inappropriate clinical
management or treatment. The external oncologist
review of the content of reports has provided laboratories
with feedback as to the positive and negative elements
within their reports and will aid in the improvement in
genomic test result reporting.
Genetics/Inherited Conditions
P001. The UHN NGS Panel for Hematologic
Malignancies: Use for Standard of Care Testing of
Acute Myeloid Leukemia in Ontario/Canada
J.M. Capo-chichi1,2,3, S. Grenier1, C. Badduke2, K. Chin1,
A. Swindell1, N. Stickle1, G. Downs1, B. Nwachukwu1, T.
Zhang2, S. Kamel-Reid1,2,3, T. Stockley1,2,3
1Genome Diagnostics, Department of Clinical Laboratory
Genetics, Laboratory Medicine Program, University
Health Network, Toronto, ON, Canada, 2Advanced
Molecular Diagnostics Laboratory, Princess Margaret
Cancer Centre, University Health Network, Toronto, ON,
Canada, 3Department of Laboratory Medicine and
Pathobiology, University of Toronto, Toronto, ON,
Introduction: Hematologic malignancies including
myeloproliferative neoplasms, myelodysplastic
syndromes and acute leukemia are characterized by
great clinical and genetic heterogeneity. Genomic
profiling is an important component for diagnosis,
prognosis and management of patients with these
hematopoietic malignancies in the updated 2016 WHO
classification of myeloproliferative neoplasms and acute
leukemia. AMP recently recommended molecular testing
of 34 critical genes as standard of care in chronic myeloid
neoplasms. Our previous work on 1,600 patients with
various hematopoietic malignancies (Princess Margaret
Cancer Center - Advanced Genomics in Leukemia study)
S16 The Journal of Molecular Diagnostics
proved that comprehensive Next-Generation Sequencing
(NGS) was an effective aid for disease diagnosis and
clinical management of these individuals.
Methods: We developed a custom hybrid-capture NGS
panel that targets 28 hotspot gene regions and 21 entire
consensus DNA coding sequences of 49 clinically
relevant genes in hematologic malignancies.
Results: This clinical test is currently offered as part of a
newly funded program for patients with acute leukemia in
Ontario/Canada. To date, we have completed molecular
profiling on over 500 Acute Myeloid Leukemia (AML)
patients. Specific approaches were taken to enable
capture of complex gene regions and to identify
challenging variants including larger insertions, deletions
and duplications in genes such as CEBPA, CALR, FLT3
and KMT2A. We have reported over 1,000 clinically
relevant variants (preliminary data on 300 patients) using
this NGS assay: minimum: 0, maximum: 13, average 4
per case. NGS detected unsuspected variants in genes
(BRAF, NOTCH1) indicating complex clinical
presentation of disease; comprehensive NGS testing was
also useful in confirmation of diagnosis of AML and in
identification of AML patients that were further confirmed
to have familial predisposition to myeloid neoplasms
(DDX41, RUNX1, MPL mutations).
Conclusions: Molecular profiling by NGS is an effective
strategy for proper diagnosis and clinical management of
patients with hematologic malignancies. Important
considerations for NGS panel testing development and
validation in a clinical setting include relevance of
hematologic malignancy genes and/or gene regions to be
tested, complete gene coverage and uniform coverage of
targeted gene regions. Bioinformatics support is required
to improve assay performance and analytical sensitivity
particularly for detection of large insertions, deletions and
P002. Real-World Diagnostic Value of Targeted Panel
Sequencing in Korean Hearing Loss Patients
J.H. Rim1, J. Jung1, H.Y. Gee1, S.-T. Lee1, J.Y. Choi1,
J.R. Choi1
1Yonsei University College of Medicine, Seoul, Republic
of Korea
Introduction: As more genes have been identified as
linked to hearing loss and more public databases
become available, a substantial proportion of patients
with hearing loss are diagnosed for genetic causes using
next generation sequencing (NGS) results. In this study,
we present clinical and diagnostic utility of a targeted
NGS panel in Korean patients with hearing loss, along
with several strategies to maximize the diagnostic yield.
Methods: A total of 39 unrelated Korean pediatric
patients with moderate to severe nonsyndromic hearing
loss (NSHL) at Severance Hospital were tested using a
customized NGS panel (Celemics, Korea) targeting all
coding exons and exon/intron boundaries of 182
deafness-related genes. After target enrichment,
sequencing was performed on the MiSeq System
(Illumina, USA) according to the manufacturer's
instructions. Our customized bioinformatics pipeline with
segregation studies, extensive literature review, and well-
curated genotype-phenotype correlation were applied to
improve the diagnostic yield. The classification of variants
followed the NSHL-optimized American College of
Medical Genetics and Genomics (ACMG) guidelines.
Results: Among 39 patients, 11 patients (28.2%) had
pathogenic or likely pathogenic mutations in 12 genes
including TECTA, POU3F4, POU4F3, NLRP3,
MYO7A, GJB2 and KCNQ4. Types of pathogenic
variants were variable including missense (n=9),
frameshift (n=3), nonsense (n=1), splice site (n=1), and
deletion (n=1). Interestingly, a 22-year old female patient
had dual pathogenic mutations in GJB2 and KCNQ4,
highlighting the clinical utility of NGS. In addition, whole
gene deletions including CATSPER2 with varying sizes
confirmed by Multiplex Ligation-dependent Probe
Amplification were identified in three patients, including
possibly two carrier individuals. Clinical genotype-based
phenotype review confirmed that the identified
pathogenic variants were strongly matched in most
patients according to NSHL-optimized ACMG guideline
Conclusions: Our results demonstrate that this targeted
NGS panel has an excellent diagnostic performance and
great utility in patients with hearing loss. Deep
understanding of the genetic architecture of hearing loss
was also of utmost importance, and thus the targeted
NGS panel could be used in routine clinical diagnostics
with careful interpretation and optimal strategies.
(Keywords: Hearing loss, Next generation sequencing,
Targeted gene panel, Deafness)
P003. Distinct Craniofacial and Skeletal Features,
Keratosis, Hearing Loss, Short Stature, and Complex
Congenital Heart Disease with Vasculopathy:
Expanding the Phenotype of BRD4-Related
L. Starr1, A. Yetman1
1University of Nebraska Medical Center, Children's
Hospital, Pediatrics, Omaha, NE, United States
Introduction: Patients with point mutations and deletions
encompassing BRD4 have been described.
Haploinsufficiency of this gene can cause a Cornelia de
Lange-like (CdL) syndrome with short stature,
microcephaly, characteristic facial features, congenital
heart defects, hirsutism, and neurodevelopmental delays.
Herein is a detailed description of a new patient with
distinct physical characteristics. An updated genetics
consultation was requested for a 29 year old male that
resulted in a gestalt of a hybrid CdL and Rubenstein
Taybi phenotype.
Methods: An exome analysis with trio reference was
performed which resulted in a single genetic mutation of
interest. A missense mutation in BRD4 (NM_058243.2,
c.860G>T (p.Glyc287Val)) was classified as that of
uncertain significance. This de novo mutation had a
minor allele frequency of zero, was highly conserved, and
was predicted in most models to be deleterious.
Results: The distinct physical characteristics included a
S17 The Journal of Molecular Diagnostics
craniofacial appearance reminiscent of the
aforementioned syndromes. He had microcephaly,
micrognathia, poor dentition, thick and course hair with
multiple whorls and hirsutism with atypical hair placement
and density. He had a very narrow chest wall. There was
a single testicle s/p orchiopexy, small hands and feet,
hyperkeratosis of the palms, scleroderma-like tight skin,
and an angiofibromatous persistent rash-like appearance
on the face. He had short stature and hearing loss,
presumed to be secondary to cholesteatomas. He had a
small adrenal mass of no known consequence. He had
intellectual disability with a pleasant personality.
Cardiovascular complications included tetralogy of Fallot
type double outlet right ventricle, progressive dilation of
the ascending aorta and moderate dilation of the celiac
Conclusions: This patient had features of CdL similar to
previously reported patients with BRD4
haploinsufficiency. The three individuals previously
reported with missense mutations were in a cohort
chosen for their CdL phenotype, which was qualified as
mild. In contrast, our patient had more distinct features.
While his moderate aortic dilation may be at least in part
due to his congenital cardiac defect, the celiac artery
dilation would support a more systemic vasculopathy.
This patient may represent the severe-end of the BRD4-
related disorder spectrum and expands the phenotype.
P004. Somatic Mosaic Truncating Mutations in Exon
6 of PPM1D in Leukocytes Are Likely to Be Related to
the Effects of Chemotherapy
D. Won1, B. Kim1, S.T. Lee1, J.R. Choi1
1Yonsei University College of Medicine, Department of
Laboratory Medicine, Seoul, Republic of Korea
Introduction: PPM1D (Protein phosphatase magnesium-
dependent 1δ) plays a key role in the negative regulation
of p53 and modulates the DNA damage response
pathway. Gain-of-function mutations of the PPM1D gene
are associated with a wide variety of cancers including
breast cancer and ovarian cancer while over suppressing
the function of p53. Reportedly, truncating mutations in
exon 6 of PPM1D can cause overexpression of PPM1D
and the mutations of mosaic form in peripheral blood are
more common in the ovarian cancer or breast cancer
group than in the control group. On the other hand, it is
reported that PPM1D mosaic truncating mutations in
leukocytes are associated with chemotherapy rather than
predisposition of solid cancers.
Methods: We examined the PPM1D mutations in next-
generation sequencing (NGS) results of patients with
suspected hereditary solid tumors. The NGS tests were
conducted on hereditary cancer panel with peripheral
blood between July 2016 and August 2018. Detected
variants were further examined by visual verification
using the Integrative Genomic Viewer (IGV). The
pathogenicity of variants is classified according to the
ACMG criteria. We also examined the mutations in blood
and cancer specimen by sanger sequencing in three
Results: A total of 1195 patients were included in the
study. Among 1,195 patients, the proportion of patients
with breast and/or ovarian cancer (80.3%) was the
highest. Truncating mutations in PPM1D were detected
in 4 (4/965, 0.41%) of the patients with breast cancer
and/or ovarian cancer. All truncating mutations were in
exon 6, in mosaic form. The mean percentage of mutant
reads was 11.15% (range= 5.4%-15.4%). All four had a
history of chemotherapy. Of the total, 395 were treated
with chemotherapy and odds ratio of chemotherapy
about PPM1D truncating mutations in exon 6 was 18.4
(95% CI: 0.99-342.67, p=0.051). In three cases, A low
percentage of mutations in peripheral blood by NGS were
confirmed by Sanger sequencing but no corresponding
mutations in the tumor tissues were observed.
Conclusions: In our NGS analysis, PPM1D mosaic
truncating mutations in peripheral blood were observed in
0.41% of breast and/or cancer patients, which was
similar to that reported previously. However, the
mutations were somewhat significant in the group treated
with chemotherapy. Clonal hematopoiesis which is
related to age cause hematologic cancers according to
acquisition of somatic mutations due to chemotherapy.
Patients with truncating PPM1D mutations in our study
may be subjects with a so-called clonal hematopoiesis of
indeterminate potential (CHIP). Nevertheless, it is unclear
whether the mosaic truncating mutations of PPM1D in
the blood are the cause acting as a driver of solid cancer
or the result of chemotherapy. We need to study more
samples and make sure that the mutations are in normal
P005. Molecular Diagnosis of Huntington's Disease in
Suspected Cases of Indian Origin
P. Chheda1, T. Dama1, D. Goradia1, S. Pande1, S.
1Metropolis Healthcare Ltd., Molecular Pathology,
Mumbai, India
Introduction: Huntington's disease (HD), an autosomal-
dominant, progressive, neurodegenerative genetic
disorder is caused by an increase in the number of CAG
repeats in the huntingtin (HTT) gene. Genetic tests that
accurately determine the number of CAG repeats are
performed for confirmation of diagnosis. The aim of our
study was to evaluate utility of triplet-primed polymerase
chain reaction (TP-PCR) for routine diagnosis of HD in
suspected cases from India.
Methods: We ran a combination of CAG flanking PCR
and triplet-primed PCR for estimation of CAG repeats in
732 cases with clinical suspicion of HD. Age of onset
versus the number of CAG repeats were studied.
Results: There were 363 cases (49.6%) that showed the
presence of expanded alleles, with 349 (47.67%) being
fully penetrant alleles and 13 (1.77%) in the reduced
penetrance category. The remaining suspected cases
(N=369) showed CAG repeats in normal range, thereby
excluding the diagnosis of HD and these cases should be
further evaluated for Huntington's disease-like (HDL)
syndromes. There were 13 juvenile cases with an age of
onset lower than 20 years, with the longest allele
comprising 106 CAG repeats found in an 8-year-old male
S18 The Journal of Molecular Diagnostics
patient. The results demonstrated an inverse (R = - 0.66)
relationship between CAG length and age at clinical
Conclusion: Our study on pan-Indian cases is one of the
largest studies reported so far in India and focuses on the
most accurate and comprehensive molecular diagnostic
evaluation of HD.
P006. Development of a Clinical Next Generation
Sequencing Panel for Diagnosis of Cystic Lung
Diseases: Evaluation of Diagnostic Yield and
Optimization of Bioinformatics Pipelines
J. Oh1, B. Kim1, Y. Kim1, K.-A. Lee1
1Gangnam Severance Hospital, Yonsei College of
Medicine, Laboratory Medicine, Seoul, Republic of Korea
Introduction: Multiple cystic lung disease (CLD)
represents a diverse group of uncommon disorders that
can present a diagnostic challenge due to the increasing
number of diseases associated with this presentation.
Among CLD, several diseases have well-defined
causative mutations in the relevant genes; e.g.
lymphangioleiomyomatosis (LAM), Birt-Hogg-Dube
syndrome (BHD), tuberous sclerosis complex (TSC) and
cystic fibrosis (CF). Thus far, the molecular diagnosis of
CLD is mainly based on Sanger sequencing. As Sanger
sequencing of all the candidate genes substantially
increase the cost, genetic testing usually starts with the
most commonly involved genes and proceeds to less
likely genes only when clinical suspicion is very high. In
recent years, targeted next-generation sequencing (NGS)
platform has been further developed, allowing us to focus
specifically on genomic regions of interest for cheaper
multiplexed sequencing of more cases. However,
inaccuracy in detecting the length of homopolymers
repeats and complexity in detecting structural variation
became a critical barrier against accurate detection of
genomic variations. Herein we seek to evaluate
diagnostic yield of customized CLD panel using NGS
Methods: 49 patients with multiple lung cysts was
enrolled. Mutations in FLCN gene were characterized
through Sanger sequencing, Multiplex Ligation-
dependent Probe Amplification (MLPA) and quantitative
PCR (qPCR). Evaluation of diagnostic yield of the CLD
NGS panel was done using Ion torrent S5 NGS platform.
Genomic DNA reference materials obtained from Coriell
cell repository and results of Sanger sequencing-
confirmed mutations in FLCN was used to validate
analytical performance of the CLD panel. Two
bioinformatics(BI) pipeline for processing NGS data were
used; the Torrent Suite(TSS) with a plug-in Torrent
Variant caller (Thermo Fisher Scientific, MA, USA) and
NextGENe(NG) software (Softgenetics, PA, USA).
Results: Out of 49 patients with multiple lung cysts, 18
FLCN pathogenic variants and 1 FLCN large deletion
were observed confirmed by Sanger sequencing and
MLPA respectively. Using CLD NGS panel test, 1 extra
pathogenic variant and 3 VOUs were newly observed. In
analytical performance evaluation, analytical sensitivity of
TSS and NG was 77.1% (59.4-89.0) and 94.3% (79.5-
99.0). Specificity of TSS and NG was both 100%. After
adapting additional hotspot bed file to TSS, sensitivity
went up to 100%. And after adapting IGV visual
inspection of the pathogenic variant hotspot list,
sensitivity of NG was 100%.
Conclusions: Diagnostic yield using NGS went up from
38.7% to 46.9% compared with FLCN Sanger
sequencing alone. Even though the pathogenic hotspot
of FLCN is mostly filtered out using TSS BI pipeline due
to 8 homopolymers repeat sequences, adjusting BI can
dramatically improve the overall performance.
Optimization of the BI pipeline is essential when
designing difficult NGS panel.
P007. DNA Methylation and Its Correlation with
Breast Cancer Pathological Prognostic Staging
V. Gupta1,2, P. Deshpande3, A. Bhake4
1Jawaharlal Nehru Medical College, Datta Meghe
Institute of Medical Sciences, Pathology, Wardha, India,
2Datta Meghe Institute of Medical Sciences, Molecular
Biology and Epidemiology Laboratory, Wardha, India,
3University of Dayton, Ohio, OH, United States,
4Jawaharlal Nehru Medical College, Pathology, Wardha,
Introduction: Breast cancer is the most common form of
malignant disease in women worldwide and is the major
cause of mortality among women. It is a heterogeneous
disease with histopathological, genetic and epigenetic
characteristics. Epigenetic alterations such as in DNA
methylation, characterized by addition of a methyl group
at the carbon-5 position of cytosine residue within CpG
dinucleotide, are considered an early event in tumor
development. However, the changes in DNA methylation
and its correlation with the progression of breast cancer
is not well studied. The study aims to assess the
relationship of DNA methylation in breast cancers
presenting at different pathological prognostic stages, for
understanding the progressive behavior.
Methods: Cross sectional study on surgically resected
mastectomy specimens from breast carcinomas,
diagnosed as invasive ductal carcinoma and matched
adjacent normal tissues from 63 breast cancer patients
were included in the study, conducted in Department of
Pathology and Molecular Biology and Epidemiology
Laboratory at Jawaharlal Nehru Medical College, Datta
Meghe Institute of Medical Sciences, Wardha, India.
Clinico-pathological data was obtained from the record in
the surgical pathological section. The specimens were
collected after pathological examination in phosphate
buffer saline.
DNA extraction and methylation: A conventional
phenol/chloroform extraction procedure was followed.
DNA methylation at CpG islands was studied using
ELISA by Epigentek kit.
Statistical analysis: performed using SPSS 23.0. Fisher's
exact tests was used to calculate the agreement from
different sources (tumor, and adjacent tissues) among
Results: Sixty-three, breast cancer patients were
enrolled in the study. Median age 50 years, range (26 -
S19 The Journal of Molecular Diagnostics
77). Molecular subtyping was as follows: Luminal A&B,
18(29%); HER2 enriched 27(42%); triple negative,
18(29%) cases respectively. Maximum number of
patients were in stage III i.e., 36 (57%) followed by stage
II: 18 (29%) and stage I: 9 (14%), as categorized in
Pathological Prognostic Staging system.
The median levels of CpG island methylation for invasive
ductal carcinoma tumor samples was 64% and that of
normal samples was 18%. There was a significant rise in
CpG island methylation from stage I to stage II i.e., 26%
to 97%. This rise could be due to increase in number of
CpG islands in tumor suppressor genes with the
increasing stage of breast cancer. This rise was also
significant after controlling the age of patients (p-value <
0.05). No significant association was found between
different molecular subtypes and levels of CpG island
methylations in DNA.
Conclusion: Aberrant DNA methylation is associated
with increased stage of breast cancer and can be used to
assess the biological behavior of tumor progression.
Status of DNA methylation at CpG islands can also be
used as a prognostic marker.
P008. Evaluation of Developmental Delay and Related
Disorders in Pediatric Patients Using Targeted Next
Generation Sequencing
J.-Y. Han1, K.-S. Woo1, I.-H. Jeong1, R.-Y. Goh1, J.-H.
1Dong-A University College of Medicine, Laboratory
Medicine, Busan, Republic of Korea, 2Dong-A University
College of Medicine, Pediatrics, Busan, Republic of
Introduction: The etiology of developmental delay (DD)
and related disorders is heterogeneous and is therefore
still posing diagnostic challenges. With the introduction of
next generation sequencing (NGS) techniques, diagnosis
and discovery of causative genes have increased
tremendously in the past several years. In this
presentation, we conducted targeted NGS using various
panels for diagnosis and classification of pediatric DD
and related conditions.
Methods: The patients were evaluated first by clinical
examination at Dong-A University hospital and further
enrolled for NGS analysis in the form of targeted
sequencing. The panels employed were hypothyroidism
panel, proportionate short stature panel, rasopathy panel,
skeletal dysplasia panel, congenital adrenal hyperplasia
(CAH) panel, and inborn error of metabolism panel,
based on the primary clinical and laboratory features of
the patients.
Results: Molecular diagnosis was performed for 25
pediatric patients. Eight proportionate short stature
panels, 7 hypothyroidism panels, 3 each of CAH and
rasopathy panels and 2 each of inborn error of
metabolism and skeletal dysplasia panels were used. Six
pathogenic variants related to the phenotype of patients
were identified by mostly hypothyroidism panel in 5
patients. The other patient showed a pathogenic variant
of FGFR3 gene with skeletal dysplasia panel. Most
common variants were obtained with DUOX2 gene.
Interestingly 13 patients showed variants of uncertain
significance (VUS).
Conclusions: DD patients show variable phenotypes
along with genetic heterogeneity. NGS analysis could be
the best choice for detection of molecular defects.
However, reevaluation of genotype-phenotype data with
literature review is still recommended to revise and
update the mutation profile especially in patients with
P009. Impact of CYP4F2 rs2108622 Variant on
Warfarin Dose Requirement in Indian Patients
S. Shah1, M. Paradkar1, S. Natarajan1, C. Ponde2, R.
Rajani2, F. Jijina3, R. Gursahani4, T. Ashavaid5
1P. D. Hinduja Hospital & MRC, Research Laboratories,
Mumbai, India, 2P. D. Hinduja Hospital & MRC,
Department of Cardiology, Mumbai, India, 3P. D. Hinduja
Hospital & MRC, Department of Clinical Hematology,
Mumbai, India, 4P. D. Hinduja Hospital & MRC,
Department of Neurology, Mumbai, India, 5P. D. Hinduja
Hospital & MRC, Department of Laboratory Medicine,
Mumbai, India
Introduction: Warfarin dose management is challenging
due to its narrow therapeutic index and large inter-patient
variability. CYP2C9, VKORC1 variants along with known
non-genetic factors accounts for ~50% of warfarin dose
variability. Additionally a CYP4F2 variant has shown to
influence warfarin dose and therefore, due to absence of
data from the western India, the present study was aimed
to determine the allele frequency as well as to determine
the impact of CYP4F2 rs2108622 variant in warfarin dose
management in Indian patients.
Methods: The present ongoing study included a total of
101 patients undergoing warfarin therapy and 30 healthy
controls. Their mean daily warfarin dose, international
normalized ratio (INR) and demographics were recorded.
CYP2C9 and VKORC1 allele frequency data was
available from our previously published data. CYP4F2
variant genotyping was performed by allele-specific PCR
method and validated by Sanger sequencing. CYP4F2
allele frequencies and genotype frequencies were
calculated and tested for Hardy-Weinberg equilibrium.
Krushal-Wallis and Chi-square test were performed for
comparison of warfarin dose and INR with CYP4F2
mutants using Graphpad Prism (v7.02). A p value < 0.05
was considered significant.
Results: The CYP4F2 rs2108622 minor allele frequency
in the patient and controls were found to be 40% and
42% respectively, which was observed to be higher than
that reported in the Caucasian population (30%). The
mean warfarin doses as well as the mean INR were not
statistically significant with CYP4F2 genotypes. However,
on subgroup analysis, it was revealed that the CYP4F2
variant allele was significantly (p< 0.05) associated in
patients with supra and sub-therapeutic INR (>3 and < 2)
as compared to patients with therapeutic INR (2-3). Most
importantly, only 14% of supra and sub-therapeutic INR
was explained by CYP2C9 and VKORC1 genotyping,
which was significantly increased to 79% with inclusion of
CYP4F2 genotyping. Further in the above subgroup, 68%
S20 The Journal of Molecular Diagnostics
of patients with CYP2C9+VKORC1 wild-type showed the
presence of CYP4F2 variant. Similar association
between CYP4F2 variant and supra and sub-therapeutic
warfarin dose was observed.
Conclusions: The present study, a first from Western
India showed the CYP4F2 rs2108622 variant to be
clinically relevant and would impact warfarin dose
management. Further validation on a large Indian cohort
will be required in order to establish the true clinical utility
of CYP4F2 variant in management of patients on
Warfarin therapy.
P010. Frequency of the Moyamoya-Related RNF213
p.Arg4810Lys Variant in Patients with Moyamoya
M.-N. Lee1, E.H. Lee2
1Green Cross Labs, Molecular Unit, Yongin-si, Republic
of Korea, 2Green Cross Labs, Yongin-si, Republic of
Introduction: Moyamoya disease (MMD) is a rare,
genetically heterogeneous cerebrovascular disease
resulting from progressive steno-occlusion of the distal
internal carotid arteries accompanied by abnormal
collateral vessels. RNF213 was identified as a
susceptibility gene for MMD and a variant in the Ring
Finger 213 gene (RNF213), altering arginine to lysine at
the codon of 4810 (c.14429G>A, p.R4810K)
(rs112735431) is associated with MMD in Asian
populations. Interestingly, many studies have reported
that a certain proportion of the general population in
Japan, Korea, and China also has this variant. Patients
with the homozygote of the variant were also reported to
present more severe form of Moyamoya disease. In this
study, we investigated the frequency of the variant
including heterozygous and homozygous frequency in
Korean MMD patients.
Methods: During January 2017 to October 2018, a total
of 460 MMD patients were included in this study. MMD
was diagnosed as either definite (bilateral) or probable
(unilateral) according to published guidelines. We
sequenced RNF213 exon 60 (encoding p.R4810K) to
determine the genotype of the variant. After obtaining
written informed consent from all subjects, peripheral
blood samples were obtained. Genomic DNA was
extracted from peripheral blood leukocytes by using
Chemagic DNA blood kit (PerkinElmer, Baesweiler,
Germany), according to the manufacturer´s instructions.
Exons of the RNF213 and their anking introns were
amplified using primer sets designed by the authors
(available upon request). The PCR was performed with a
thermal cycler (model 2720, Applied Biosystems, Foster
City, CA, USA). Direct sequencing was performed with
the BigDye Terminator Cycle Sequencing Ready
Reaction kit (Applied Biosystems) on the AB 3730 DNA
analyzer (Applied Biosystems).
Results: The c.14429G>A (p.R4810K) variant was
identified in 44.3% (204/460) of patients with definite or
probable MMD. A heterozygote of the variant was
observed in 43.0% (198/460) patients and homozygote
was identified in 1.3% (6/460) patients. Homozygotes
had a slightly earlier age at diagnosis compared with
heterozygotes or wild types (median age at Diagnosis 43,
49 and 52 yrs, respectively), which did not achieve the
statistical significance. However, the proportion of
childhood onset (age under 15 yrs) were significantly
higher in patients with homozygotes compared with
patients with heterozygotes.
Conclusions: In summary, our results confirm that
alteration in RNF213 predisposes patents of Korean
MMD patients and the homozygous c.14429G>A
(p.R4810K) variant in RNF213 could be a useful
molecular marker for predicting the severe type of MMD,
for which proper medical intervention is recommended,
and may provide a better monitoring and prevention
strategy. This study will provide the frequency of
c.14429G>A (p.R4810K) variant in RNF213, and an
effective strategy for the molecular diagnosis of MMD in
P011. Thiopurine S-Methyltransferase (TPMT) Allele
Frequencies in Hong Kong Population
S.K.F. Cheung1, J.C.Y. Ho1, Y.S. Chan1, I. Tang1, J.
1Queen Mary Hospital, Division of Transplantation and
Immunogenetics, Department of Pathology, Hong Kong,
Hong Kong
Introduction: Thiopurine drugs, such as Azathioprine, 6-
Mercaptopurine and 6-Thioguanine, are effective
immunosuppressive agents in the management of
inflammatory bowel disease, autoimmune diseases and
organ transplantation. It is also used in anti-leukemia
therapy. Two modes of action have been proposed for
thiopurine. Firstly, the thiopurine metabolite, 6-methyl
Thioinosine monophosphate, inhibits the cell proliferation
of fast growing lymphocytes by suppressing de novo
purine synthesis. In addition, 6-thioguanine nucleotide,
another thiopurine metabolites, exerts a pro-apoptotic
effect to activated T cells [1]. Thiopurine S-
methyltransferase (TPMT) is the key enzyme responsible
for catalytic conversion of the thiopurine metabolites to
non-cytotoxic products. TPMT variants with reduced
enzyme activity may cause accumulation of these
cytotoxic active thiopurine metabolites which can lead to
myelosuppression [2]. Studies have shown that the
TPMT gene is more polymorphic in the Western
populations than in the Chinese populations [3-5]. In this
study, we investigated the allele frequencies of the most
common TPMT variants (TPMT*2, *3A, *3B and *3C) in
the Hong Kong population.
Methods: Genomic DNA was extracted from whole blood
of 300 Hong Kong healthy blood donors. The TPMT*2,
*3B and *3C variant alleles were first detected using a
modified Amplification-Refractory Mutation System
(ARMS) PCR method [6]. Beta-2-microglobulin was used
as the internal control. All samples with variants detected
were further confirmed by Sanger sequencing of the
TPMT exon V, VII and X. Rare TPMT alleles were not
Results: TPMT*1/*3C genotype was detected and
confirmed in 9 out of 300 healthy subjects. TPMT*2 and
S21 The Journal of Molecular Diagnostics
TPMT*3B alleles were not detected. The allele frequency
of TPMT*3C allele in the Hong Kong population is 1.5%.
Conclusions: This study showed the allele frequency of
the most common TPMT variants in Hong Kong. Hong
Kong as well as the Chinese population has a lower
frequency of TPMT*3C variant as compared with the
Western populations and only heterozygous TPMT*1/*3C
genotype (3%) was found. It was estimated that 3% of
the Hong Kong population has an intermediate TPMT
enzyme activity. TPMT genotype could be performed for
patients requiring thiopurine therapy to guide the
reduction of drug dose to minimize the risk of
P012. Characterization and Distribution of Dystrophin
Gene Mutations Observed in Patients Suspected of
Duchenne Muscular Dystrophy in Pakistan
Z. Ahmed1, A. Nasir1
1Aga Khan University, Pathology and Laboratory
Medicine, Karachi, Pakistan
Introduction: Duchenne muscular dystrophy (DMD) is
an inherited X-linked recessive genetic disorder caused
by a defect in the dystrophin gene. This results in
quantitative and/or qualitative abnormality in the
dystrophin protein causing progressive muscular damage
and weakness. The objective of this study was to
investigate the spectrum of deletions and duplications in
the dystrophin gene of patients suspected of DMD or of
being DMD carriers.
Methods: A retrospective review of DMD testing
performed from March 2017 to June 2018 was conducted
at Aga Khan University Hospital, Karachi. MLPA was
used for identification of deletion/duplication in 79 exons
of the dystrophin gene.
Results: In total 92 cases, 85 males and 7 Females were
examined. In males, deletions and duplication were
identified in 44 (52%) of the 85 cases, deletions were
found in 38 cases (45%) and duplications in 6 cases
(7%). In 7 cases referred for females, 2 of the 7 females
(30%) were found to have a heterozygous deletion,
suggestive of carrier status.
Conclusions: This study identifies deletions in
dystrophin exon 42-52 to be the most common while
duplications were mostly found in exons 3-7. This
information can facilitate the use of new treatments with
exon-skipping drugs, which are mutation specific for
P013. Association of Vitamin D Receptor (VDR) Start
Codon FokI Polymorphism with Acute Myelogenous
M. Faiz1, A. Shahid1
1Institute of Nuclear Medicine and Oncology, Lahore,
Introduction: The role of chromosomal translocations,
gene mutations, and polymorphisms in the pathogenesis
of leukemia/lymphoma have been extensively studied.
Recently, the role of vitamin D and vitamin D receptor
(VDR) polymorphisms in hematologic malignancies has
been considered. Vitamin D start codon polymorphism
has been suggested to be associated with an increased
risk of developing different hematological malignancies
including AML. The objective of this study was to
investigate the association between VDR start codon F/f
polymorphism and the risk of AML among Pakistani
Methods: A total of 70 Pakistani patients with acute
myelogenous leukemia were enrolled in this hospital-
based case-control study. In addition, 70 age and sex
matched healthy volunteers were included as a control
group. Five milliliters (ml) of venous blood were collected
and genomic DNA was extracted from all samples using
standard methods. This was followed by the detection of
VDR Fok-I polymorphism by polymerase chain reaction
and restriction fragment length polymorphism.
Results: A 260 bp fragment of the VDR gene was
amplified by PCR and digested by restriction enzyme
(FokI) to detect the polymorphism. The genotype FF was
the most frequent (76%) in patients with AML, followed
by the genotype Ff (20%) and genotype ff(4%). Similarly
in the control group the genotype F/F also was the most
frequent (57%) followed by the genotype f/f (39%) and F/f
genotype (4%). There was statistically significant
association between AML and each of the genotypes FF
and Ff (p=0.0001). No statistically significant correlation
between the VDR Fok-I polymorphism and sex/age
(P=0.611) was found.
Conclusion: Both FF and Ff genotypes are associated
with increased risk of AML among Pakistani patients.
P014. Human Genome Reference Assembly for Blood
Group Systems
D.-H. Ko1, J.J. Yang1, H. Kim2, S.-H. Hwang1, H.-B. Oh1
1Asan Medical Center, Department of Laboratory
Medicine, Seoul, Republic of Korea, 2Seoul National
University Hospital, Department of Laboratory Medicine,
Seoul, Republic of Korea
Introduction: Clinical applications of next-generation
sequencing (NGS) are continuously expanding. Human
reference genome build comparison is a commonly used
method for variant detection but such an approach is
limited for polymorphic regions encoding blood group
antigens. In this study, we analyzed and typed the
sequences of two human reference genomes to provide
a reference blood group for application of NGS.
Methods: Two human reference genome builds,
GRCh37 and GRCh38 were analyzed. Representative
reference allele sequences were selected for clinical
important blood groups: ABO, Rhesus, Kidd, and Kell
blood groups (AJ536122.1 for ABO gene; X63094.1 and
L08429.1 for RHD gene; DQ322275.1 for RHCE gene;
NM_015865.7 for SLC14A1 gene; and M64934 for KEL
gene). The reference alleles were ABO*A1.01, RHD*01,
RHCE*01, JK*02, and KEL*02, respectively. Sequences
from the two genome builds were compared to the
reference alleles to detect the presence of variants in
coding regions.
Results: Compared with the reference ABO allele, both
reference genome builds had c.261delG, representative
S22 The Journal of Molecular Diagnostics
of the O allele. In addition, GRCh37 assembly showed
additional variants in coding regions (c.106C>T,
c.188G>A, c.189C>T, and c.220C>T). Multiple intronic
variants were found in both reference genome builds. For
RHD, there was no difference between the two reference
genome builds. On the other hand, c.636C>A and
c.1036T>C variants were detected when compared to the
L08429.1 and X63094.1, respectively. These findings are
possibly sequencing artifacts within the reference alleles
and require further analysis. There was no difference
between the reference alleles and the referenced
genome builds for RHCE and SLC14A1. Several intronic
differences were found between the two reference builds
for KEL.
Conclusions: Several differences between the reference
genome build and reference alleles were found. Two
reference genome builds also had differences. Findings
from our study suggest that caution is necessary when
using reference genome builds for polymorphic regions
such as blood group antigens.
P015. An Artificial Intelligence Engine for High-
Throughput Matching of Genetic Variants to Their
ACMG/AMP Classification for Inherited Disease Gene
E. Frise1, S. Nohzadeh-Malakshah1, M. Falcioni1, C.
Bellavia1, F.M. De La Vega1,2, E. Kiruluta1
1Fabric Genomics Inc., Oakland, CA, United States,
2Stanford University School of Medicine, Dept of
Biomedical Data Science, Stanford, CA, United States
Introduction: The ACMG and AMP defined a set of
evidence-based guidelines to support variant
pathogenicity assessment and reporting in inherited
disease diagnostics. The guidelines define several
criteria, each assessing particular supporting evidence
information that is checked independently for each
variant. Criteria are grouped by different levels of
evidence and a set of rules combines the evaluated
criteria and classifies a variant accordingly as pathogenic
(P), likely pathogenic (LP), benign (B), likely benign (LB),
or uncertain significance (VUS). Although widely adopted
in clinical interpretation of inherited disease variants, this
process has remained a largely manual and time-
consuming process in clinical labs that is not scalable to
the volume of NGS testing data. The rich evidence
needed to properly classify variants is difficult to
effectively incorporate in manual classification schemes,
resulting in discrepancies of interpretation, excessive
specialized reviewing labor, and higher costs. Informatics
tools have been proposed to ease the application of the
guidelines in clinical practice, such as the web-based
ClinGen Pathogenicity Calculator. However, such tools
do not solve the issue of interpreting a large set of
variants, since they do not automate the entire
classification process.
Methods: For these reasons, we developed
VariantMatch, an artificial intelligence engine to
automatically infer the classification of variants that
utilizes a forward-chaining inference engine at its core to
implement the ACMG-AMP criteria returning a predicted
classification for each variant. The engine is able to
incorporate criteria and rule refinements for specific
genes and related diseases by separating the rule
specification from the code that executes them. A critical
input is careful curation of the gene-disease relationship
attributes and computed features including variant
deleteriousness scores such as VVP. Moreover, a natural
language generation module provides a rationale of the
classification reached for reference by clinical geneticists
when reviewing and reporting cases. To evaluate the
performance of our method, we analyzed a set of 2,978
of BRCA1 variants of diverse prior classifications
compiled from ClinVar, VariSNP, JapanData, and the
Color Database.
Results: Our preliminary results show that VariantMatch
was able to correctly classify automatically 96% of 2 stars
or more ClinVar variants and was able to reclassify 69%
and 76% of Color and JapanData VUS variants. The
output provided explanatory text indicating the key rules
matched that led to a classification and was easily
understandable by clinicians.
Conclusion: We plan to integrate VariantMatch into
Fabric Enterprise, a cloud-based software platform for
clinical genomic analysis and reporting, to enable high-
throughput genetic labs to score variants rapidly,
reproducibly, cost-effectively and at scale.
P016. Development of a NGS-based Neotelomere
Mapping Method for Detection of Terminal Deletions
B. Zhang1, A. Iqbal1, W. Crowe1, P. Elliott1, A. Campbell1,
A. Perkins1, P. Rothberg1
1University of Rochester Medical Center, Pathology,
Rochester, NY, United States
Introduction: Human telomeres are composed of
repetitive (TTAGGG)n sequences. Broken chromosomes,
when not repaired by homologous recombination or non-
homologous end-joining, can be stabilized by telomere
capture or by the telomerase-mediated direct addition of
telomeric repeats to nontelomeric DNA ends, forming a
neotelomere (Flint J et al., 1994 PMID:7521575). This
end-repair process is called telomere healing, which
results in terminal deletions. Neotelomeres are found in
human terminal deletion syndromes, which occurs in at
least 1/5,000 individuals (Shaffer LG et al., 2000
PMID:11092830; Knight SJL et al., 1999
PMID:10568569). However, it is not easily detected by
current assay designs and informatics pipelines, and has
not been characterized in a systematic manner.
Methods: We developed a novel NextGen-sequencing
(NGS) based method and an informatics algorithm to
map neotelomeres attributable to telomere healing. This
method uses one telomere-anchored primer and one
ligated adaptor primer together for library generation
followed by deep sequencing. We applied this assay in a
pilot study to human samples with known terminal
deletions, and successfully identified telomere junctions
for most chromosomes in the initial 4 specimens as well
as disease-causing terminal deletions.
Results: The clinical performance and utility of the assay
to screen for fetal chromosomal terminal deletions in cell-
S23 The Journal of Molecular Diagnostics
free DNA from maternal plasma are being assessed, and
Conclusion: We are also developing a cellular model to
study the molecular pathways regulating neotelomere-
mediated end repair.
P017. Determining Genetic Predispositions Using
Ultrafast Amplicon-Based NGS Hereditary Cancer
L. Lee1, Y.L. Liu1, K. Pendleton1, L. Lin1, G. Liu1, J. Liu1,
Z. Liu1
1Paragon Genomics, Inc., R&D, Hayward, CA, United
Introduction: As genetic testing becomes more widely
accepted, medical professionals are increasingly
interested to use it as a means to screen for hereditary
conditions. While hybrid capture-based target enrichment
methods are mainstream for surveying large number of
targets, they are time-consuming and cumbersome to
perform requiring highly trained operators with
specialized equipment.
Methods: We have developed a multiplex PCR-based
CleanPlex technology to make it possible to produce
high-quality coverage of genomic targets using a fast and
sample target enrichment and NGS library construction
workflow. Here we present an NGS panel targeting
hereditary mutations that covers the full exons of 37
genes using 1445 pairs of PCR primers. In addition, the
panel also detects two clinically relevant intronic
mutations. This panel paired with the CleanPlex workflow
can enable the quick analysis of genes associated with
an increased risk of developing hereditary cancers using
small amounts of genomic DNA.
The CleanPlex 3-step workflow includes a multiplex PCR
step with targeted primers, a background cleaning step to
remove by-products, and a final PCR to add Illumina
adapter sequences and sample indexes. NGS libraries
were made using 10 ng of genomic DNA per pool (40 ng
total). The CleanPlex Hereditary Cancer Panel contains
primers divided into four pools for the multiplex PCR.
Samples were sequenced at ~2,500 read depth on an
Illumina NextSeq. Sequenced reads were demultiplexed.
Mapping rate and on-target rates were calculated, and
variants were identified using Paragon Genomics' variant
calling algorithm.
Results: The samples exhibit >96% uniformity at 0.2X
mean, and all exons were sufficiently covered to provide
accurate variant calling information. The assay is highly
reproducible and sensitive, exhibiting an R2 value of 0.95
between independently prepared replicates, and a
detection rate of >99% for single nucleotide variants.
Conclusion:The Paragon Genomics CleanPlex
Hereditary Cancer Panel can be used for rapid profiling
of cancer risk by using small quantities of genomic DNA.
With the easy workflow and quick turnaround, this panel
can make genetic monitoring easy, efficient, and
P018. Understanding Next Generation Sequencing
for Autosomal Dominant Polycystic Kidney Disease
H. Rennert1, A. Tan1, J. Blumenfeld1
1Weill Cornell Medical College, New York, NY, United
Introduction: ADPKD is a genetically heterogeneous
disorder caused by mutations in PKD1 and PKD2,
accounting for ~85% and 15% of cases, respectively.
PKD1 consists of 46 exons spanning ~52 Kb of genomic
DNA, a large part of which is duplicated in six homologue
genes (i.e. pseudogenes) showing ~98% homology with
PKD1 exons 1-33. PKD2 spans 15 exons and a coding
sequence of ~3Kb. Gene-based mutation testing
currently is the predominant method of ADPKD
genotyping. However, the marked allelic heterogeneity of
the disease-associated mutations, the vast majority of
which are private, and the duplicated structure of PKD1
together with its large size, make mutation screening a
significant technical challenge.
Methods: The key step in ADPKD genotyping
procedures is selective amplification of PKD1 sequences,
while excluding the pseudogenes. This was traditionally
achieved by utilizing long-range PCR (LR-PCR) with
primers located to the rare mismatch sites that
distinguish PKD1 and the pseudogenes, while the single-
copy regions of PKD1 and PKD2 are directly amplified
from gDNA. Amplicons are then analyzed by Sanger
sequencing. Introduction of next generation sequencing
(NGS) has revolutionized the field of molecular genetics
and was recently applied to PKD genetic testing. In this
method, LR-PCR is coupled with NGS. Bar-coded LR-
PCR libraries from individual patients are pooled together
and analyzed using paired-end NGS in a single
sequencer flow-cell (Illumina). Sequencing results are
then sorted according to the sample barcode, aligned
against the reference PKD genes and mutations are
called using a computational bioinformatics analysis
Results: This approach has a sensitivity of 100%
compared with Sanger sequencing, while retaining high
specificity, enabling identification of single nucleotide
variants (SNVs), indel and copy number variation (CNV).
Moreover, the high read-depth and average coverage
(>400x) obtained over exonic regions significantly
increases testing sensitivity compared to Sanger
sequencing, allowing detection of low-level mutations
(~3% VAF), especially important for the detection of
genetic mosaicism. Recently, both WES and WGS have
been applied to PKD gene testing, eliminating the need
for LR-PCR. However, the sensitivity of these methods
varied significantly between the duplicated and single-
copy PKD1 compared to LR-PCR based sequencing
(30%~80% versus 100% sensitivity), primarily due to low
read-depth particularly in GC-rich regions and mapping of
reads from duplicated regions.
Conclusion: NGS has the potential to dramatically
improve PKD genetic testing by simplifying workflow and
allowing simultaneous detection of SNV/indels and CNV
S24 The Journal of Molecular Diagnostics
in a single test, increasing mutation detection rate and
significantly reducing costs and turn-around time.
However, careful evaluation of these novel methods and
the corresponding computational pipelines is required to
ensure high quality clinical testing.
P019. A New Scalable and Automatable Method for
the Extraction of cfDNA
J. Doh1
1Beckman Coulter Life Sciences, Genomic Reagents,
Indianapolis, IN, United States
Introduction: Cell free DNA (cfDNA) consists of small
(150 - 500 bp) DNA fragments that circulate in the blood.
cfDNA levels are generally low in healthy, non-pregnant
patients, and increase in patients with cancer, pregnancy,
or extensive damage to tissue. cfDNA is believe to derive
mostly from apoptotic cells, and biomarkers for a variety
of diseases have been found in cfDNA, increasing
interest in its use as a less invasive way to monitor
disease progression.
Methods: In this study, cfDNA was extracted from
horizon control and cancer patient samples. The
mechanism of DNA extraction was optimized to maximize
yield of cfDNA. The DNA extraction is scalable, and able
to work with low amounts of plasma (~200 µL) up to
several milliliters of plasma. The DNA extraction can also
be run on a variety of automation, and was run on both
Kingfisher and Biomek machines. The efficacy of this
extraction method was measured by looking at detection
of known mutations in horizon control samples and the
effect of input amount on detection levels was
determined. Cancer patient samples were also extracted
and analyzed for associated mutations.
P020. Assessment of Minimum Residual Disease
Detection in Multiple Myeloma Samples Using Next
Generation Sequencing Based LymphoTrack Assays
and Flow Cytometry
Y. Huang1, A. Jacobsen1, J. Panganiban1, E. Vigil1, K.
Hutt1, F.W. Martinez1, M. Blankfard1, A. Medina2, C.
Jiménez2, R. García-Sanz2, J.E. Miller1
1Invivoscribe, San Diego, CA, United States, 2Hospital
Universitario de Salamanca-IBSAL, Salamanca, Spain
Introduction: Multiple myeloma (MM), characterized by
the presence of >10% clonal plasma cells in bone
morrow (BM), represents the second most common
hematological malignancy. Multiparameter flow cytometry
(MFC) is a standard tool used to detect and monitor MM
patients. Next generation sequencing (NGS) based
methods have demonstrated advantages with improved
sensitivity, and international organizations (NCCN, IMWG
and ESMO) have recently included NGS for minimum
residual disease (MRD) assessment in MM. Here we
report the results of a pilot study comparing the
performance of NGS-based LymphoTrack Assays and
MFC by testing 101 MRD MM samples.
Methods: 101 paired BM samples from MM patients
were tested in this study. The MFC methods utilize the 8-
color direct immunofluorescence technique and test 1-5
million cells. Our NGS based MRD assessment tests only
~1/10 as many cell equivalents of genomic DNA (0.7µg,
~107,000 cell equivalents) from the collected specimens
which were blinded prior to testing with four
LymphoTrack Assays (IGH FR1, FR2, FR3 and IGK) on
a MiSeq. LymphoTrack Software was used to sort data
by target and index. The clonal rearrangement
(clonotype) identified by LymphoTrack FR1 or FR2 Assay
was then tracked to test the 91 subsequent samples.
LymphoQuant Internal Control was added to each PCR
reaction at 100 cell equivalents to estimate the cell
equivalents within each sample. LymphoTrack Software -
MiSeq and LymphoTrack MRD software were used to
analyze the sequencing results from baseline and follow
up samples, respectively.
Results: Despite the limited DNA available for testing,
out of 101 baseline samples, 84 (83%), 80 (79%), 63
(62%) and 87 (86%) samples were detected as clonal
positive by IGH FR1, FR2, FR3 and IGK, respectively.
When combing FR1 and FR2, 94% (95/101) clonal
positivity was achieved. When combining all 4 targets,
100% (101/101) clonal positivity was achieved.
Using LymphoTrack MRD Software to identify clonotype
sequences, we performed MRD assessment on 91
samples that were positive for IGH FR1 or FR2. 47 (52%)
were positive, 43 (47%) were negative, and 1 (1%) was
invalid, due to insufficient reads. MRD results by MFC on
the same 91 patients showed 45 (49%) positive, 37
(41%) negative, and 9 (10%) invalid samples. Excluding
samples with DNA less than 700 ng (~107,000 cell
equivalents), concordance was 85.4%. Three samples
were detected by MCF but not by NGS while four
samples were detected by NGS but not by MCF.
Conclusions: LymphoTrack assays were shown to
detect clonal sequences in 100% of MM baseline
samples and, despite testing approximately 1/10 the
number of cell equivalents, were able to achieve 85.4%
agreement with MFC in detecting MRD. This suggests
that LymphoTrack assays are a useful tool in identifying
and tracking disease status in MM samples. Unlike MFC,
the LymphoTrack assays and accompanying
bioinformatics software can be submitted for approval to
regulatory authorities worldwide.
S25 The Journal of Molecular Diagnostics
P021. Comparative Analysis of Molecular
Approaches Used for Targeted Detection of
Mutations in FLT3 and NPM1 from DNA and RNA
C. Wei1,2,3, J.P. Nadeau1, S. Grenier1, B. Nwachukwu1,
J.M. Capo-Chichi1,2,3, S. Kamel-Reid1,2,3, T. Stockley1,2,3
1Genome Diagnostics, Department of Clinical Laboratory
Genetics, Laboratory Medicine Program, University
Health Network, Toronto, ON, Canada, 2Advanced
Molecular Diagnostics Laboratory, Princess Margaret
Cancer Centre, University Health Network, Toronto, ON,
Canada, 3Department of Laboratory Medicine and
Pathobiology, University of Toronto, Toronto, ON,
Introduction: FLT3 and NPM1 are key biomarkers in
diagnosis, prognosis and disease management of Acute
Myeloid Leukemia (AML). WHO recommends molecular
profiling of Internal Tandem Duplications in FLT3 (FLT3-
ITD), hotspot missense variants in the Tyrosine Kinase
Domain of FLT3 (FLT3-TKD) and insertions in NPM1 for
standard of care testing in AML. Molecular approaches
used to characterize FLT3-ITD, FLT3-TKD and NPM1
mutations differ in performance and analytical sensitivity;
there is no consensus on whether levels of FLT3-ITD
should be assessed from DNA or RNA templates.
Methods: We developed fluorescent PCR based
fragment analysis for targeted detection of FLT3-ITDs,
FLT3-TKDs and NPM1 insertions from DNA and RNA; in
parallel, we used a Next-Generation Sequencing (NGS)
panel for comprehensive analysis of 49 genes altered in
hematologic malignancies, which include FLT3 and
NPM1. We compared the performance of NGS, DNA and
RNA-based fragment analyses approaches on peripheral
blood or bone marrow from 100 AML patients. We
assessed the diagnostic yield of these different
methodologies; variant allelic fraction from DNA versus
RNA, length of FLT3-ITD captured, assay sensitivity.
Assay limitations were also evaluated in this study.
Results: Preliminary data on 50 patients shows that
genotyping calls for FLT3-ITD, FLT3-TKD and NPM1
insertions are concordant in 47/50 of cases analyzed.
Discordances were seen for variant with allelic fraction
detected below 5% by fragment analysis and missed by
NGS. Allelic fraction of variants, particularly FLT3-ITDs
differ between DNA and RNA templates; allelic fractions
were also inconsistent when comparing NGS and
fragment analysis results performed on DNA. NGS was
effective in resolving the nature of in-frame and out of
frame insertions/duplications; amplicon sizing using
fragment analysis could not differentiate these types of
variants occurring in FLT3.
Conclusion: NGS and fragment analysis are
complementary strategies for detection of FLT3-ITDs,
FLT3-TKDs and NPM1 insertions. Fragment analysis is
suitable for critical clinical decision making in AML
patients due to faster turnaround time; NGS
comprehensive molecular profiling aids with diagnosis,
prognosis and management of AML. Recent reports are
in favor of measurements of FLT3-ITDs by RNA; further
investigations and correlation of FLT3-ITDs levels with
clinical outcome of patients will be useful in determining
the optimal nucleic acid template (DNA or RNA) to use
for better stratification of AML patients.
P022. Clinical Application of Next-Generation
Sequencing in Myeloproliferative Neoplasms:
Genetic Profiles and Prognostic Significance for
Personalized Medicine
J.M. Lee1, E. Han1, K. Han1, M. Kim1, Y. Kim1
1Catholic University of Korea, Laboratory Medicine,
Seoul, Republic of Korea
Introduction: The mutational landscape of
myeloproliferative neoplasms (MPNs) has been
discovered including three diagnostic driver mutations,
JAK2, MPL, and CALR. Nowadays, next-generation
sequencing (NGS), especially targeted panel
sequencing, is widely applied in clinical laboratory to
identify mutational profile in hematologic malignancies.
Herein, we evaluated genetic profiles in MPNs using
NGS and explored the clinical application in the single
center cohort.
Methods: A total of 432 patients were included; 196
essential thrombocythemia (45.37%), 112 polycythemia
vera (25.9%), 72 primary myelofibrosis (16.7%), 14 other
MPNs (3.2%) and 38 myelodysplastic/myeloproliferative
neoplasms (MDS/MPN, 8.8%). JAK2 V617F, CALR and
MPL 515 mutations were firstly analyzed using allele-
specific PCR and fragment analysis. Targeted panel
sequencing including 86 genes was performed. We
analyzed relationships between genetic proflies and
clinical outcomes including acute transformation, bone
marrow fibrosis and death.
Results: Three driver mutations were detected in 82.4%
of patients; JAK2 V617F in 64.8%, CALR in 15.7% and
MPL in 1.9%. We also detected mutations in other genes
through NGS which included ASXL1 (27.0%), TET2
(7.6%), KMT2C (3.2%), SRSF2 (2.9%), RUNX1 (2.9%),
CSF3R (2.5%), ATM (2.2%), SETBP1 (1.8%), DNMT3A
(1.8%), TP53 (1.8%), U2AF2 (1.8%), KMT2D (1.8%) and
PTPN11 (1.8%). Among the three phenotypic subtypes,
PMF showed shorter event free survival (EFS) than the
others (hazard ratio: PMF-ET 2.62, PMF-PV and 2.97). In
the aspect of genetic profile, TP53 mutation showed
worst outcome followed by chromatin or spliceosome
mutation (Median EFS: 77 and 99 months, 95%
Confidential interval: 5.0-264 and 72-226, respectively).
Conclusions: We demonstrated that NGS effectively
identified mutations that were more relevant to the
prognosis in MPN patients. Genetic profiles can improve
a risk stratification to make efficient therapeutic plan
including targeted therapy in MPN patients.
S26 The Journal of Molecular Diagnostics
P023. Potential Role of JAK2V617F and ASXL1
Mutations as Biomarkers for Disease Outcomes
Amongst Thai Patients with Philadelphia-Negative
Myeloproliferative Neoplasms
P. Arnutti1, D. Apipongrat2, T. Numbenjapon2, T.
Nimmanon1, P. Jindatanmanusan3
1Phramongkutklao College of Medicine, Pathology,
Bangkok, Thailand, 2Phramongkutklao College of
Medicine, Medicine, Bangkok, Thailand,
3Phramongkutklao College of Medicine, Pediatric,
Bangkok, Thailand
Introduction: Polycythemiavera (PV), essential
thrombocythemia (ET), and primary myelofibrosis (PMF)
are three main disorders constituting Philadelphia-
negative myeloproliferative neoplasms (Ph-negative
MPNs). The JAK2V617F mutation is present in the
majority of these patients and has been included in
diagnostic criteria for these conditions. Furthermore,
ASXL1 mutations have been demonstrated as prognostic
markers for risk stratification of MPNs. This study
therefore aimed to determine the prevalence of
JAK2V617F and ASXL1 mutations and their relationship
with disease outcomes in Thai Ph-negative MPNs.
Methods: A total of 139 cases (78 male and 61 female)
of Ph-negative MPNs, consisting of 51 PV cases, 67 ET
cases, 18 PMF cases, and 3 MPN-U cases, were
enrolled. All cases were investigated for JAK2V617F and
ASXL1 mutations, utilising allele-specific PCR (AS-PCR)
and Sanger sequencing, respectively.
Results: The JAK2V617F mutation was detected in
64.8% of Ph-negative MPN cases (72.6% of PVcases,
61.2% of ET cases, 50.0% of PMF cases, and 100.0% of
MPN-U cases), whereas ASXL1 mutations were detected
in 10.8% of Ph-negative MPN cases (22.2% of PMF
cases, 11.8% of PV cases, 7.5% of ET cases, and 0.0%
of MPN-U cases). ASXL1 mutations were demonstrated
in 7 out of 90 JAK2V617F-negative cases (7.8%) and 8
out of 49 JAK2V617F-positive cases (16.3%).
Interestingly, JAK2V617-negative ASXL1-positive
patients were shown to have distinctive clinical profiles in
all groups of patients. However, no difference in
thrombosis-free survival was seen amongst MPN
patients with different mutations.
Conclusions: JAK2V617F and ASXL1 mutations might
become potential biomarkers for disease outcomes in
Thai patients with Ph-negative MPNs.
P024. Accurate Detection of Low AF Variants
Relevant to AML by Anchored Multiplex PCR and
Next Generation Sequencing
V. Johnson1, K. Moore1, L. Griffin1, A. Berlin1, A. Licon1,
R. Walters1
1ArcherDX, Boulder, CO, United States
Introduction: Acute Myeloid Leukemia (AML) is clinically
and biologically heterogeneous, requiring the detection of
multiple mutations for characterization. For instance,
FLT3-ITDs and CEBPA mutations represent important
markers in AML, however they are difficult to detect by
NGS due to the highly variable nature of ITDs, the high
GC content of CEBPA, and the difficulty in mapping
repeated sequences to a wild-type reference. Tracking
low frequency mutations is also of growing importance.
The ability to accurately detect variants at low allele
fractions (AFs) using a single test can be used to assess
treatment efficacy and potential relapse.
Methods: We developed Archer VariantPlex myeloid
assays based on Anchored Multiplex PCR (AMP) to
detect important mutations in myeloid malignancies. AMP
is a target enrichment strategy that uses molecular-
barcoded adapters and gene-specific primers for
amplification, permitting open-ended capture of DNA
fragments from a single end. This approach enables
flexible and strand-specific primer design to provide
better coverage of ITD-containing and GC-rich regions.
We also developed a method to assess SNV sensitivity
taking into account both unique coverage depth and
noise for single base substitutions. This strategy enables
utilization of position-specific detection thresholds and
maximizes sensitivity and specificity. We tested this
approach using the VariantPlex Core Myeloid panel, by
titrating reference inputs into background normal samples
to examine detection of low AF variants.
Results: Our assay enables calling of a 30bp FLT3-ITD
down to sub-0.05% allele frequencies. Using optimized
low AF conditions improves coverage depth, consistency
of low AF FLT3-ITD detection, and sensitivity (98.5% of
bases are powered to call a true variant at an allele
frequency of 3.0% with 1M reads and 200ng of input).
We show >1000X unique molecule coverage across the
coding region of CEBPA and use this challenging region
to visualize the minimum detectable AF (MDAF) at which
a variant has a >95% probability of being detected above
the noise (95MDAF). Finally, we show consistent single
nucleotide variant (SNV), insertion and deletion (indel),
and ITD calling at sub-0.5% allele frequencies, and
demonstrate the utility of reporting variant-specific
MDAFs and normal dataset P-values when analyzing low
AF variants.
Conclusion: AMP provides NGS-based detection of
complex mutation types that are relevant in AML. We
demonstrate robust calling of FLT3-ITDs and other
variants at low AFs. We also demonstrate full coverage
of CEBPA with high depth and low noise such that the
95MDAF predicts confident variant calling at low AFs.
This approach is accurate and scalable, enabling
simultaneous detection of multiple mutation types across
multiple target genes in a single assay.
S27 The Journal of Molecular Diagnostics
P025. Downregulation of RAG1 Expression Inhibits
the Leukemic IK6 Isoform Formation in B-Cell
Lymphoblastic Leukemia Cells
L. Lyu1,2, W. Zhao1, T. Huang1,2, H. Tu3, P. Ru3, X. Pen3,
D. Jones3
1The Ohio State University College of Medicine,
Pathology, Columbus, OH, United States, 2Wuhan No.1
Hospital, Tongji Medical College, Urology, Wuhan, China,
3The Ohio State University College of Medicine, James
Molecular Laboratory at Polaris, Columbus, OH, United
Introduction: The IKZF1 gene isoform, IK6 is a common
pathogenic finding in acute lymphoblastic leukemia (ALL)
with or without BCR-ABL1 fusion gene. The formation of
IK6, however, might be caused by either exon deletion or
aberrant RNA splicing. RAG1, one of the key elements
for DNA V(D)J recombination, was predicted to be the
key player in a previous study ( Harvey et al. Blood 2010)
but no direct evidence was given. In this study, we
examined if down regulation of RAG1 would affect the
IK6 formation and what are the other effects of RAG1 in
the biology of ALL.
Methods: The RAG1 gene was specifically knocked
down by small interference RNA (SiRAG1) or
nonspecifically down regulated by givinostat, a type II
Histone Deacetylase and potent apoptosis inducer for
ALL (Li and Yao et al., 2016 and 2017). The presence
IK6 isoform was measured by quantitative RT-PCR. In
addition, a next generation sequencing (NGS) was
performed to study whole transcription analysis to
investigate the differentially expressed whole RNA in the
down-regulated RAG1 ALL.
Results: The IK6 isoform was only detected in RNA but
not in genomic DNA from this cell line confirmed to be
resulted aberrant RNA splicing. The RAG1 was
significantly knocked down by both by 200 nM SiRAG1
and 25 uM Givinostat, 65% and 90%, respectively, (P<
0.001). Meanwhile, IK6 was co-inhibited, 60% and 75%,
respectively (P < 0.001) compared to the controls while
wild-type IKZF1 was not significantly affected. Further
WTS study showed that knockdown RAG1 only
upregulate SOX11 and CCND2 genes and downregulate
most CDK genes associated with cell cycle regulation.
Finally, the analysis of 150 upregulated genes by
SiRAG1 were positively correlated to givinostat treated
specimen (P=0.016), indicating these genes might be
involved in givinostat induced antineoplastic effect
including the apoptosis.
Conclusion: The role of RAG1 in RNA splicing
regulation has not been studied and our results in this
study were the first to support this hypothesis. In addition,
our results provided the evidence that the formation of
IK6 in B-cell ALL can be regulated by RAG1 or
expression of RAG1 with deacetylation inhibitor,
givinostat. Our finding not only revealed the molecular
mechanism in IK6 isoform formation but provided
evidence of clinical investigation if RAG1 might be a
target for personalized medicine in treatment of acute
lymphoblastic leukemia.
P026. Panel-Based Next-Generation Sequencing of
Chinese Diffuse Large B-Cell Lymphoma
B. Cao1,2, C. Sun1,2, R. Bi1,2, Z. Liu1,2, Y. Jia1,2, W. Cui1,2,
M. Sun1,2, X. Zhou1,2
1Fudan University Shanghai Cancer Center, Department
of Pathology, Shanghai, China, 2Shanghai Medical
College, Fudan University, Department of Oncology,
Shanghai, China
Introduction: Diffuse large B-cell lymphoma (DLBCL),
an aggressive and heterogeneous malignancy, remains a
challenging clinical problem, as up to one-third of patients
are not cured with initial therapy. Recently, next-
generation sequencing (NGS) studies in DLBCL have
uncovered recurrent gene mutations. We designed a
NGS multi-gene panel to identify genetic characteristics
of Chinese DLBCL patients and provide relevant
information for panel-based NGS tests to clinical
Methods: A panel of 116 DLBCL genes was designed,
based on literature, the Catalogue of Somatic Mutations
in Cancer (COSMIC) database, and FoundationOne
Heme. We studied 96 Chinese DLBCL biopsy samples
using targeted sequencing.
Results: We identified the mutation frequency of SPEN
(17%) and DDX3X (6%) in our study was higher than
Western studies. Despite functional similarities between
NOTCH1 and NOTCH2, there was no overlap in cases of
NOTCH1 and NOTCH2 mutations, suggesting a distinct
pathogenesis of N1 (NOTCH1 mutations) subtype and
N2 (NOTCH2 mutations) subtype. KMT2D pathogenic
mutations in total patients and patients treated with R-
CHOP-like therapy were associated with significantly
favorable prognosis. MYD88 L265P mutation, MYD88
pathogenic mutations, TP53 and BCL2 pathogenic
mutations were important influence factors of poor
prognosis in DLBCL.
Conclusions: This study underlines the genetic
heterogeneity of DLBCL and demonstrates the
contribution of panel-based NGS to molecular targeted
therapy in DLBCL.
P027. Low-Frequency Variant Detection in Cell-Free
DNA with Ultrafast Amplicon-Based Targeted
Sequencing Using Unique Molecular Identifiers
Y.L. Liu1, K. Pendleton1, L. Lee1, L. Lin1, G. Liu1, J. Liu1,
Z. Liu1
1Paragon Genomics, Inc., R&D, Hayward, CA, United
Introduction: The use of cell-free DNA (cfDNA)
applications has seen tremendous growth in recent years
with liquid biopsy being a noninvasive and easily
obtainable sample type with several diagnostics and
prognostic values. While liquid biopsy can potentially
enable applications such as early cancer detection,
treatment monitoring, and drug resistance screening, it
also presents new challenges to variant detection due to
the low fraction of mutant DNA in cfDNA and artifacts
from background noise resulting from PCR and
sequencing errors. While some success has been made
S28 The Journal of Molecular Diagnostics
by using adapters carrying unique molecular identifiers
(UMIs) in hybrid capture-based methods and performing
deep sequencing, such approaches suffer from long,
tricky, and tedious workflows and disappointingly low on-
target rates.
Methods: We have developed the CleanPlex UMI
technology to provide a fast, simple, and reliable NGS
solution to low-frequency variant detection. The
technology features three simple steps to generate
molecular-barcoded and target-enriched NGS libraries in
3.5 hours. This amplicon-based method consists of first,
a multiplex PCR reaction for molecular barcoding and
target amplification, then followed by a proprietary
background removal step, and finally a second round of
PCR to add platform-specific adapter sequences and
sample indexes. We validated this technology with an
NGS panel optimized for cfDNA to target hotspots in 23
genes frequently mutated in lung cancer. NGS libraries
were prepared from commercial reference cfDNA at
minor allele frequencies (mAF) ranging from 0.1% to
0.5% and sequenced on an Illumina NextSeq to evaluate
the performance.
Results: Overall, the detection sensitivity for low-
frequency alleles is high even at low DNA inputs. We
observe significant reductions in the number of false
positive calls when analyzing the samples using UMIs.
Nearly all known mutations in the reference cfDNA are
detected at 0.1% mAF. At an mAF of 0.25%, all known
mutants are detected with 100% PPV using 50ng of
DNA. Detection of all mutants at mAFs of 0.5% and PPV
of 100% require only 20ng or less of DNA. We are also
able to accurately quantify the detected allele frequencies
due to the high numbers of reference and mutant UMIs
identified for each genetic target.
Conclusion:The CleanPlex UMI technology
demonstrates high sensitivity for the detection of low-
frequency alleles with low false positive detection rates.
P028. Targeted Sequencing Reveals Pathogenic
Genes in EBV-associated T/Natural Killer-Cell
Lymphoproliferative Disorders
L.-M. Gao1, W.-P. Liu1
1West China Hospital of Sichuan University, Pathology,
Chengdu, China
Introduction: EBV associated T/NK cell
lymphoproliferative disorders are characterized by clonal
proliferation of EBV positive T or NK cells. Extranodal
NK/T cell lymphoma-nasal type (ENKTL) is the most
common type of this diseases and it can occur both in the
upper aerodigestive track (nasal-ENKTL) and extranasal
sites (extranasal-ENKTL). However, diverse biological
behaviors and prognosis were observed for ENKTL
involving different sites. The mechanism underlying such
heterogeneity remains elusive. Aggressive natural killer
cell leukemia (ANKL) and EBV positive T cell
lymphoproliferative disorders (EBV+T-LPD) are relatively
rare. There are some overlapping clinicopathologic
features among ENKTL, ANKL and EBV+T-LPD.
Methods: A total of 169 cases of EBV associated T/NK
cell lymphoproliferative disorders were collected. This
included 123 ENKTL cases (87 nasal-ENKTL and 36
extra-nasal ENKTL), 12 ANKL cases and 34 EBV+T-
LPD cases. The targeted sequencing gene panel covers
all exons and elected introns of 64 genes involving in the
development of lymphoma. The mutation profiles of
ENKTL, ANKL and EBV+T-LPD were compared.
ENKTL is the most common disease of the three
EBV+T/NK-LPD, we further done the analysis of gene
mutation of ENKTL cases, Comparison of mutated genes
between nasal and extra-nasal ENKTL, protein express
detection of KMT2D, TET2, EP300 and NOTCH1,
relationship analysis of protein loss expression and gene
mutation and prognosis in 123 target sequencing ENKTL
Results: STAT3, KMT2D, DDX3X, NOTCH1 and TET2
showed high mutation rates in these three EBV+T/NK-
LPD. The comparison of mutated genes of these three
disorders showed that the exclusive mutated genes of
ENKTL are PRDM1 et al., the ANKL are ITPKB et al., the
EBV+T-LPD is MFHAS1. The mutated rates of KMT2D
and NOTCH2 in extra-nasal ENKTL are greater than
those in nasal-ENKTL.
Further study in ENKTL showed that the loss of KMT2D
and TET2 protein were associated with the mutation of
KMT2D and TET2 gene respectively. Prognostic analysis
showed that the mutation of KMT2D or TET2 gene, the
loss of expression of KMT2D or TET2 protein all
significantly correlated with inferior overall survival of
ENKTL cases.
Conclusion: We conclude that the exclusive mutated
genes of ENKTL, ANKL and EBV+T-LPD are different
the mutation of KMT2D and TET2 may play an important
role in the development of ENKTL.
P029. Longitudinal Monitoring of AML Tumors with
High-Throughput Single-Cell DNA Sequencing
Reveals Rare Clones Prognostic for Disease
Progression and Therapy Response
D. Eastburn1, C. McMahon2, R. Durruthy-Durruthy1, M.
Carroll2, C. Smith3, A. Perl2
1Mission Bio, Inc., South San Francisco, CA, United
States, 2University of Pennsylvania, Division of
Hematology-Oncology, Philadelphia, PA, United States,
3University of California, San Francisco, Division of
Hematology-Oncology, San Francisco, CA, United States
Introduction: AML (acute myeloid leukemia) is
increasingly being treated with precision medicine. To
better inform treatment, the mutational content of patient
samples must be accurately determined. However,
current tumor sequencing paradigms are inadequate to
fully characterize many instances of the disease. A major
challenge has been the unambiguous identification of
potentially rare and genetically heterogeneous neoplastic
cell populations, capable of critically impacting tumor
evolution and the acquisition of therapeutic resistance.
Standard bulk population sequencing is unable to identify