© 2010 Nature America, Inc. All rights reserved.
Nature GeNetics ADVANCE ONLINE PUBLICATION
The per-generation mutation rate in humans is high.
De novo mutations may compensate for allele loss due to
severely reduced fecundity in common neurodevelopmental
and psychiatric diseases, explaining a major paradox
in evolutionary genetic theory. Here we used a family
based exome sequencing approach to test this de novo
mutation hypothesis in ten individuals with unexplained
mental retardation. We identified and validated unique
non-synonymous de novo mutations in nine genes. Six of
these, identified in six different individuals, are likely to be
pathogenic based on gene function, evolutionary conservation
and mutation impact. Our findings provide strong experimental
support for a de novo paradigm for mental retardation.
Together with de novo copy number variation, de novo point
mutations of large effect could explain the majority of all
mental retardation cases in the population.
Recent studies1,2 have indicated that humans have an exceptionally
high per-generation mutation rate of between 7.6 × 10−9 and 2.2 ×
10−8. An average newborn is calculated to have acquired 50 to 100
new mutations in his or her genome, resulting in approximately 0.86
new amino-acid–altering mutations2. Spontaneous germline muta-
tions can have serious phenotypic consequences when they affect
functionally relevant bases in the genome. In fact, their occurrence
may explain why diseases with a severely reduced fecundity remain
frequent in the human population, especially when the mutational
target is large and comprised of many genes. This would explain a
major paradox in the evolutionary genetic theory of mental dis-
orders3,4. In agreement with this hypothesis, de novo copy number
variations (CNVs) are a known cause of schizophrenia, autism and
mental retardation5,6. Much less is known about the frequency and
impact of de novo point mutations in these common diseases. Whole
genome or exome sequencing now permits the study of these muta-
tions and their role in disease in a systematic genome-wide manner.
This approach has recently been used to identify causative genes in
several rare syndromes1,7–10. In addition, targeted resequencing of the
coding exons of the X chromosome revealed nine genes associated with
X-linked forms of mental retardation11, showing the strength of these
analyses in common diseases. In this study, we used a family based
whole-exome–sequencing approach to test the de novo mutation hypo-
thesis in an unselected cohort of individuals with mental retardation.
We sequenced the exomes of ten case-parent trios. All cases, eight
males and two females, had moderate to severe mental retardation
and a negative family history. Clinical evaluation did not lead to a
syndromic or etiologic diagnosis (Supplementary Note). Prior
cytogenetic analysis showed normal chromosomes, and array-based
genomic profiling did not reveal de novo or other CNVs associated
with mental retardation. In addition, fragile X syndrome was excluded
by FMR1 repeat expansion analysis. On average, we obtained 3.1 Gb
of mappable sequence data per individual after exome enrichment
(37 Mb of genomic sequence targeting ~18,000 genes) and sequenc-
ing on one quarter of a SOLiD sequencing slide (Online Methods and
Supplementary Table 1). Color space reads were mapped to the refer-
ence genome. On average, 79.6% of the bases originated from the tar-
geted exome, with 90% of the targeted exons covered at least ten times.
The median exon coverage was 42-fold, indicating that the majority
of variants present in each exome could be robustly detected using a
custom bioinformatic analysis pipeline (Supplementary Fig. 1).
On average, we identified 21,755 genetic variants per individual with
high confidence (Table 1 and Supplementary Fig. 2). We developed an
automated prioritization scheme to systematically identify all candidate
dominant de novo mutations in each affected individual (Fig. 1). We
first excluded all nongenic, intronic and synonymous variants other
than those occurring at canonical splice sites. This first step reduced
the number of candidates to an average of 5,640 non-synonymous
and canonical splice site variants per affected individual. We further
reduced this number to 143 by excluding all known, likely benign,
variants by comparison with data from dbSNP database v130 and our
in-house variant database. Next, we used the exome data from each
case’s parents to exclude all remaining inherited variants. This resulted
in an average of five (with a range of two to seven) candidate de novo
non-synonymous mutations per affected individual (Table 1).
For all 51 candidate mutations (Supplementary Table 2), we per-
formed Sanger sequencing to (i) validate the mutations observed in
the probands and (ii) validate the absence of the mutations in the
parental DNA. Thirty-eight candidates could not be validated in the
proband (covered by a median of five variant reads in the exome
sequencing experiment), but 13 candidates could be validated
A de novo paradigm for mental retardation
Lisenka E L M Vissers1,2, Joep de Ligt1,2, Christian Gilissen1, Irene Janssen1, Marloes Steehouwer1,
Petra de Vries1, Bart van Lier1, Peer Arts1, Nienke Wieskamp1, Marisol del Rosario1, Bregje W M van Bon1,
Alexander Hoischen1, Bert B A de Vries1, Han G Brunner1,3 & Joris A Veltman1,3
1Department of Human Genetics, Nijmegen Centre for Molecular Life Sciences and Institute for Genetic and Metabolic Disorders, Radboud University Nijmegen
Medical Centre, Nijmegen, The Netherlands. 2These authors contributed equally to this work. 3These authors jointly directed this work. Correspondence should be
addressed to H.G.B. (firstname.lastname@example.org) or J.A.V. (email@example.com).
Received 11 August; accepted 18 October; published online 14 November 2010; doi:10.1038/ng.712
© 2010 Nature America, Inc. All rights reserved.
ADVANCE ONLINE PUBLICATION Nature GeNetics
(covered by a median of 17 variant reads). Parental analysis validated
the de novo occurrence for 9 of these 13 mutations, detected in seven
different individuals (Table 2 and Supplementary Figs. 3 and 4).
We did not identify these mutations in a total of 1,664 control chro-
mosomes, nor did we see other likely pathogenic mutations identi-
fied in the affected genes in these control chromosomes, indicating
that the population frequency of these types of de novo mutations in
these genes will be lower than 0.22% (power = 0.95, α = 0.05). Eight
of the de novo mutations were present in a heterozygous state on
the autosomes and one was present in a hemizygous state on the
X chromosome. All de novo mutations occurred in different genes,
including two genes recently implicated in mental retardation
(Table 2). In addition to using a dominant disease model, we also
analyzed the data for recessive forms of mental retardation. In the
affected male of trio 10, we identified a maternally inherited non-
synonymous variant in JARID1C (Table 2), which is a well-described
X-linked mental retardation gene12. Subsequent analysis of this vari-
ant in DNA obtained from the affected individual’s grandparents indi-
cated that the mutation had occurred de novo in the mother of this
proband. No conclusive evidence for autosomal recessive inheritance,
either homozygous or compound heterozygous, was obtained for the
other affected individuals.
Next, we evaluated the function of each mutated gene in relation
to the disorder (Table 2). Three genes do not seem to play a role in
biological pathways linked to mental retardation. BPIL3 is involved
in the innate immune response13, whereas PGA5 is involved in pro-
tease activity in the stomach14. The function of ZNF599 is currently
unknown. For the six other genes affected by de novo mutations, func-
tional evidence suggests a role in mental retardation. Two mutations
occurred in genes (RAB39B and SYNGAP1) that, when disrupted, are
known to cause mental retardation (Table 2)15,16. For the remaining
four mutated genes, evidence for a causal link with mental retardation
is provided by model organisms and protein-protein interaction stud-
ies. DYNC1H1 encodes a cytoplasmic dynein that acts as a motor for
intracellular retrograde axonal transport. Heterozygous Dync1h1+/−
mutant mice exhibit sensory neuropathy17, and studies in zebrafish
have shown the importance of dync1h1 in correct nuclear position-
ing. Mislocalization of nuclei in the vertebrate central nervous system
is likely to result in profound patterning defects and severely com-
promised function18. Notably, DYNC1H1 interacts with PAFAH1B1,
the gene associated with type I lissencephaly, which involves gross
disorganization of the neurons within the cerebral cortex19. YY1
encodes the ubiquitously expressed transcription factor yin-yang 1
and directs histone deacetylases and histone acetyltransferases, impli-
cating chromatin remodeling as its main function. Complete abla-
tion of Yy1 in mice results in early embryonic lethality, whereas Yy1
heterozygous mice display growth retardation, neurulation defects
and brain abnormalities20. Recent studies show that YY1 inter-
acts directly with MECP2; MECP2 is mutated in Rett syndrome21.
DEAF1 encodes a transcription factor that regulates the 5-HT1A
receptor in the human brain. Mutations in the Drosophila DEAF1
ortholog result in early embryonic arrest, suggesting an essential role
for the gene in early development22. Additional evidence is provided
by Deaf1-deficient mice, which show neural tube defects including
exencephaly23. Finally, CIC is a member of the HMG-box transcrip-
tion factor superfamily, which is associated with neuronal and glial
development of the nervous system. CIC is predominantly and tran-
siently expressed in immature granule cells of the cerebellum, hippo-
campus and neocortex, suggesting a critical role in central nervous
We next examined the evolutionary conservation of affected nucleo-
tides (using the phyloP score), as well as the potential of the de novo
mutations to affect the structure or function of the resulting proteins
(using the Grantham score; Table 2). All de novo missense mutations
and the inherited X-linked mutation were included in this analysis;
no Grantham scores were available for the additional nonsense and
frameshift mutations. Of note, de novo mutations in genes with a
functional link to mental retardation showed a higher phyloP (mean,
4.7) and Grantham score (mean, 135) than mutations in genes with-
out such a functional indication (mean phyloP score, −0.5 and mean
Grantham score, 38). We also compared these scores to those for all
non-synonymous variants in the dbSNP database as well as those
in the Human Gene Mutation Database (HGMD). The distribution
of phyloP scores and Grantham scores differed markedly between
dbSNP and the HGMD (Online Methods and Supplementary Fig. 5).
The four mutations in genes functionally linked to mental retardation
all showed higher probability values for being observed in HGMD
table 1 Overview of all variants detected per proband and impact of the prioritization steps for selecting candidate non-synonymous
de novo mutations
High-confidence variant calls
After exclusion of nongenic, intronic
and synonymous variants
After exclusion of known variants
After exclusion of inherited variants
Exome data of 10 mental retardation cases
sequenced on SOLiD 3 Plus System
Read mapping and
Default mapping settings
High-stringency variant calling
Exclude low quality
Exclude nongenic, intronic and synonymous
Exclude known SNPs and in-house database
Test occurrence in control cohort
Figure 1 Experimental work flow for detecting and prioritizing sequence
variants. For all ten mental retardation trios, prioritization of variants
observed in the probands was based on selection for non-synonymous
changes of high quality only and exclusion of all variants previously
observed in healthy individuals, together with those variants that were
inherited from an unaffected parent. Interpretation of de novo variants
was based on gene function and the impact of the mutation.
© 2010 Nature America, Inc. All rights reserved.
Nature GeNetics ADVANCE ONLINE PUBLICATION
(mean, 0.83) than for being observed in dbSNP (mean, 0.17). The
three mutations in genes without a functional link to mental retar-
dation showed an average probability of 0.94 for being observed
in dbSNP and an average probability of 0.06 for being observed in
HGMD (Table 2). Additionally, the inherited JARID1C mutation
showed a probability of 1.00 for being in HGMD versus 2.09 × 10−6
for being in dbSNP.
This analysis of the mutated nucleotides and their impact on gene
function strongly supports pathogenicity for six of the nine de novo
mutations. Importantly, these six mutations occurred in genes with a
functional link to mental retardation, two of which are known mental
retardation genes. In contrast, three de novo variants in genes without
a functional link did not appear to significantly affect protein func-
tion. Moreover, we identified a maternally inherited mutation in a
known X-linked mental retardation gene that arose de novo in the
proband’s mother. Although we have not provided individual func-
tional tests to prove causality, these data collectively provide strong
evidence for a major role of de novo mutations in mental retardation.
The identification of recurrent mutations in these genes in unrelated
cases would provide additional proof for disease causality, but this
may require the evaluation of thousands of affected individuals. The
identification of subtle CNVs encompassing (part of) these genes
may also provide additional proof for disease causality, as was shown
recently for mutations in X-linked mental retardation genes25. As of
yet, no such CNVs have been reported, nor have we found such CNVs
in our diagnostic cohort of ~4,500 individuals with mental retarda-
tion (data not shown).
The discovery of nine de novo non-synonymous mutations in this
cohort of ten affected individuals is concordant with the recently
estimated background mutation rate of 0.86 amino-acid–altering
mutations per newborn in controls2, but it will be important to com-
pare this result to similar data from healthy control trios when avail-
able. Notably, after applying the same systematic filtering approach
and Sanger sequencing, we could only validate a single de novo syn-
onymous mutation, which occurred in GRIN1 (c.351C>T, seen in
trio 10). This base pair is not conserved through evolution (phyloP
score = −3.2) and does not seem to alter splicing, suggesting that
this mutation is an unlikely candidate for causing mental retarda-
tion. Of note, the individual carrying this mutation also carries the
JARID1C mutation. The observed ratio of non-synonymous to syn-
onymous de novo mutations is far greater than would be expected
for protein-coding genes under purifying selection and indicates
that many of these mutations will result in a reproductive dis-
advantage. In contrast, the average non-synonymous to synonymous
ratio reported in dbSNP for the six genes with predicted pathogenic
mutations is significantly lower than that of the three genes with
mutations reflecting the background mutation rate (Fisher’s Exact
test, P = 0.0016), which is to be expected for disease genes in the
In summary, our results suggest that de novo mutations are
a major cause of unexplained mental retardation. These muta-
tions can readily be identified using a family based exome
sequencing approach and require only limited follow-up by
Sanger sequencing. Our findings have implications for preven-
tive and diagnostic strategies in mental retardation. Systematic
genome-wide resequencing in parent-child trios may uncover
further examples of this de novo paradigm for other human
table 2 Overview of all de novo variants identified by exome sequencing in ten individuals with unexplained mental retardation
Gene Trio SexaNM number
being observed in
De novo mutations
DYNC1H11M NM_001376 c.11465A>Cp.His3822Pro 5.5 77 0.200.80Retrograde axonal transporter;
interacts with PAFAH1B1
(mutation of which causes
lissencephaly, a neurodevel-
2.65 × 10−4Unknown
– Known X-linked mental
1.00 Ubiquitously expressed
transcription factor; mouse
knockdown results in growth
retardation, neurulation defects
and brain abnormalities;
interacts with MECP2, a known
mental retardation gene
0.03 Innate immune response
0.16 Precursor of pepsin
0.99Transcription factor; regulator
of 5-HT1A receptor in the
brain; mouse knockout causes
neural tube defects
0.54Granule cell development in
central nervous system
– Known autosomal dominant
mental retardation gene
YY13MNM_003403 c.1138G>T p.Asp380Tyr6.9 160 2.27 × 10−6
X-linked inherited mutations
JARID1C 10M NM_001146702c.1919G>A p.Cys640Tyr5.1194 2.09 × 10−6
1.00 Known X-linked mental
aSex of proband, with M for male and F for female. bVisual representation of probabilities are provided in supplementary Figure 5. Grantham scores for nonsense (in RAB39B) and frameshift
mutations (in SYNGAP1) could not be calculated.
© 2010 Nature America, Inc. All rights reserved.
ADVANCE ONLINE PUBLICATION Nature GeNetics
URLs. 1000 Genomes Project, http://www.1000genomes.org; dbSNP,
http://www.ncbi.nlm.nih.gov/projects/SNP/; HGMD, http://www.
hgmd.cf.ac.uk/ac/index.php; R, http://www.r-project.org/.
Methods and any associated references are available in the online
version of the paper at http://www.nature.com/naturegenetics/.
Accession codes. The genomic reference sequence for DYNC1H1
can be found under the GenBank accession number NM_001376;
for ZNF599 under NM_001007248; for RAB39B under NM_171998;
for YY1 under NM_003403; for BPIL3 under NM_174897; for PGA5
under NM_014224; for DEAF1 under NM_021008; for CIC under
NM_015125; for SYNGAP1 under NM_006772; for JARID1C under
NM_001146702; and for GRIN1 under NM_021569.2.
Note: Supplementary information is available on the Nature Genetics website.
We thank R. de Reuver and J. Hehir-Kwa for bioinformatics support in data
analysis and personnel from the Sequencing Facility of our department for timely
completion of Sanger sequencing of validation experiments. This work was funded
in part by grants from The Netherlands Organization for Health Research and
Development (ZonMW grants 916-86-016 to L.E.L.M.V., 917-66-36 and 911-08-
025 to J.A.V. and 917-86-319 to B.B.A.d.V.), the EU-funded TECHGENE project
(Health-F5-2009-223143 to J.d.L. and J.A.V.) and the AnEUploidy project (LSHG-
CT-2006-37627 to A.H., B.W.M.v.B., H.G.B., B.B.A.d.V. and J.A.V.).
J.A.V., L.E.L.M.V. and H.G.B. conceived the project and planned the experiments.
B.B.A.d.V. and B.W.M.v.B. performed sample collection and reviewed phenotypes.
L.E.L.M.V., A.H., I.J., M.S., P.d.V., B.v.L. and P.A. performed next-generation
sequencing experiments using a custom pipeline set up by C.G. and A.H. J.d.L.
and C.G. analyzed and interpreted the data with support from N.W. and M.d.R.
L.E.L.M.V., P.d.V., I.J. and M.S. performed validation experiments. L.E.L.M.V.,
J.d.L. and J.A.V. prepared the first draft of the manuscript. All authors contributed
to the final manuscript.
CoMPEtING FINANCIAL INtEREStS
The authors declare no competing financial interests.
Published online at http://www.nature.com/naturegenetics/.
Reprints and permissions information is available online at http://npg.nature.com/
1. Roach, J.C. et al. Analysis of genetic inheritance in a family quartet by whole-
genome sequencing. Science 328, 636–639 (2010).
2. Lynch, M. Rate, molecular spectrum, and consequences of human mutation.
Proc. Natl. Acad. Sci. USA 107, 961–968 (2010).
3. Keller, M.C. & Miller, G. Resolving the paradox of common, harmful, heritable
mental disorders: which evolutionary genetic models work best? Behav. Brain Sci.
29, 385–404 (2006).
4. Uher, R. The role of genetic variation in the causation of mental illness: an evolution-
informed framework. Mol. Psychiatry 14, 1072–1082 (2009).
5. Cook, E.H. Jr. & Scherer, S.W. Copy-number variations associated with
neuropsychiatric conditions. Nature 455, 919–923 (2008).
6. de Vries, B.B. et al. Diagnostic genome profiling in mental retardation. Am. J. Hum.
Genet. 77, 606–616 (2005).
7. Ng, S.B. et al. Exome sequencing identifies the cause of a mendelian disorder.
Nat. Genet. 42, 30–35 (2010).
8. Lupski, J.R. et al. Whole-genome sequencing in a patient with Charcot-Marie-Tooth
neuropathy. N. Engl. J. Med. 362, 1181–1191 (2010).
9. Hoischen, A. et al. De novo mutations of SETBP1 cause Schinzel-Giedion syndrome.
Nat. Genet. 42, 483–485 (2010).
10. Sobreira, N.L. et al. Whole-genome sequencing of a single proband together with
linkage analysis identifies a Mendelian disease gene. PLoS Genet. 6, e1000991
11. Tarpey, P.S. et al. A systematic, large-scale resequencing screen of X-chromosome
coding exons in mental retardation. Nat. Genet. 41, 535–543 (2010).
12. Jensen, L.R. et al. Mutations in the JARID1C gene, which is involved in
transcriptional regulation and chromatin remodeling, cause X-linked mental
retardation. Am. J. Hum. Genet. 76, 227–236 (2005).
13. Mulero, J.J. et al. Three new human members of the lipid transfer/lipopolysaccharide
binding protein family (LT/LBP). Immunogenetics 54, 293–300 (2002).
14. Taggart, R.T. et al. Relationships between the human pepsinogen DNA and protein
polymorphisms. Am. J. Hum. Genet. 38, 848–854 (1986).
15. Giannandrea, M. et al. Mutations in the small GTPase gene RAB39B are responsible
for X-linked mental retardation associated with autism, epilepsy, and macrocephaly.
Am. J. Hum. Genet. 86, 185–195 (2010).
16. Hamdan, F.F. et al. Mutations in SYNGAP1 in autosomal nonsyndromic mental
retardation. N. Engl. J. Med. 360, 599–605 (2009).
17. Chen, X.J. et al. Proprioceptive sensory neuropathy in mice with a mutation in the
cytoplasmic Dynein heavy chain 1 gene. J. Neurosci. 27, 14515–14524 (2007).
18. Tsujikawa, M., Omori, Y., Biyanwila, J. & Malicki, J. Mechanism of positioning the
cell nucleus in vertebrate photoreceptors. Proc. Natl. Acad. Sci. USA 104, 14819–14824
19. Tai, C.Y., Dujardin, D.L., Faulkner, N.E. & Vallee, R.B. Role of dynein, dynactin,
and CLIP-170 interactions in LIS1 kinetochore function. J. Cell Biol. 156, 959–968
20. He, Y. & Casaccia-Bonnefil, P. The Yin and Yang of YY1 in the nervous system.
J. Neurochem. 106, 1493–1502 (2008).
21. Forlani, G. et al. The MeCP2/YY1 interaction regulates ANT1 expression at 4q35: novel
hints for Rett syndrome pathogenesis. Hum. Mol. Genet. 19, 3114–3123 (2010).
22. Veraksa, A., Kennison, J. & McGinnis, W. DEAF-1 function is essential for the early
embryonic development of Drosophila. Genesis 33, 67–76 (2002).
23. Hahm, K. et al. Defective neural tube closure and anteroposterior patterning in
mice lacking the LIM protein LMO4 or its interacting partner Deaf-1. Mol. Cell.
Biol. 24, 2074–2082 (2004).
24. Lee, C.J. et al. CIC, a member of a novel subfamily of the HMG-box superfamily,
is transiently expressed in developing granule neurons. Brain Res. Mol. Brain Res.
106, 151–156 (2002).
25. Whibley, A.C. et al. Fine-scale survey of X chromosome copy number variants and
indels underlying intellectual disability. Am. J. Hum. Genet. 87, 173–188
© 2010 Nature America, Inc. All rights reserved. Download full-text
Subjects. Ten individuals with unexplained moderate to severe mental retar-
dation (with normal karyotypes and genomic profiles obtained using 250K
SNP arrays) were selected for exome sequencing (Supplementary Note).
Family history for mental retardation was negative for all cases. Nongenic
causes for mental retardation, including pre-, peri- and post-natal infection
and perinatal injury, were excluded. DNA was obtained from peripheral blood
from the ten probands as well as from their unaffected parents. DNA isola-
tion was performed using QIAamp DNA Mini Kit (QIAGEN), according to
the instructions of the manufacturer. This study was approved by the Medical
Ethics Committee of the Radboud University Nijmegen Medical Centre, and
all participants signed written informed consent.
Library generation. Exome enrichment required 3 μg of genomic DNA, and
an AB SOLiD Optimized SureSelect Human Exome Kit (Agilent) was used for
enrichment, containing the exonic sequences of ~18,000 genes and covering a
total of ~37 Mb of genomic sequence, as specified by the company. We followed
the manufacturer’s instructions (version 1.5) for enrichment with a minor
modification, which was the reduction of the number of post-hybridization
ligation-mediated PCR cycles from 12 cycles to 9 cycles.
SOLiD sequencing. The enriched exome libraries were subsequently used for
emulsion PCRs, following the manufacturer’s instructions (Life Technologies),
based on a library concentration of 1 picomolar (pM) (version March 2010).
For each sample, one-quarter of a sequencing slide (Life Technologies) was
used on a SOLiD 3 Plus System.
Mapping of variants. Color space reads were mapped to the hg18 reference
genome with the SOLiD bioscope software v1.2, which utilizes an iterative
mapping approach. Single-nucleotide variants were subsequently called by
the diBayes algorithm26 using high stringency settings, requiring calls on each
strand. Small insertions and deletions were detected using the SOLiD Small
Indel Tool. We assumed a binomial distribution with a probability of 0.5 of
sequencing the variant allele at a heterozygous position. Under this assump-
tion, at least ten reads are required to obtain a 99% probability that at least
two reads contain the variant allele. Variants and indels were selected using
strict quality control settings, which included the presence of at least four
unique variant reads (that is, having different start sites), as well as the variant
being present in at least 15% of all reads. All called variants and indels were
combined and annotated using a custom analysis pipeline (resulting in HCDiff
files for each individual).
Custom bioinformatic analysis pipeline. All variants reported in the HCDiff
files were filtered to ensure an optimal prioritization process. For this, we
first excluded all nongenic, intronic (other than canonical splice sites) and
synonymous variants, reducing the number of variants to an average of 5,640
per individual. Second, all known variants were excluded by comparison
with data from dbSNP v130 as well as from our in-house variant database.
At the time of this study, this in-house database contained variants from
(i) 78 in-house performed ‘exomes’, contributing 515,480 variants, and
(ii) the 1000 Genomes Project (see URLs) and published data from various
other studies27–29, contributing 3,059,835 variants, thereby bringing the
number of variants in the in-house database to 3,525,278. Of note, if the variant
observed in the proband occurred at a genomic position known in dbSNP v130,
but the change present was different in the two (for example, A/C in dbSNP
but A/T in the proband), the variant was not excluded from analysis. The fil-
tering step using this data further reduced the average number of variants to
143 per proband.
Next, for a dominant model of disease, we used the exome data from accom-
panying parents to exclude all inherited variants. This step further reduced the
number of potential de novo variants to an average of 33 per proband. As not
all variants identified in the exomes of the probands may have been sequenced
at sufficient coverage in the parental samples, we checked all remaining vari-
ants in the exome data from the accompanying parents. In brief, even if only
a single read showed the variant allele in one of the parental exome samples,
the variant was excluded for validation in the proband. Simultaneously, we
checked all remaining potential de novo indels for annotation differences in
each child-parent trio and excluded those that were found to be identical vari-
ants in both parent and child. After this final check, an average of five potential
de novo variants per proband remained for further validation.
To evaluate the presence of recessive mutations, variant filtering was
essentially performed as described above, with the main difference being that
uniquely inherited parental variants were not excluded here. The remaining
variants were evaluated for the presence of compound heterozygous variants,
as well as variants that were present in >80% of all reads. Subsequently, parental
exome data were used for segregation analysis of the variants identified.
dbSNP and HGMD. To explore the pathogenicity of our de novo variants, the
genomic evolutionary conservation score (phyloP) and the amino-acid change
(Grantham) were compared to those scores present in dbSNP (build 130) and
the HGMD (see URLs). All non-synonymous changes reported in dbSNP and
HGMD were retrieved, and overlap between databases was removed from both
datasets. In addition, non-synonymous variants in dbSNP with an OMIM
disease entry, suggestive for a Mendelian phenotype, were omitted from the
Next, quadratic discriminant analysis30 was performed on these two datasets
to determine the significance of the phyloP and Grantham scores as discrimi-
nating factors. Statistical tests were performed using the R statistics package
(see URLs). The assumption of normality in the data required for the model
was determined using Lilliefors (Kolmogorov-Smirnov) normality testing31:
PhyloP D = 0.0626, P < 2.2 × 10−16; Grantham D = 0.0828, P < 2.2 × 10−16;
PhyloP × Grantham D = 0.1395, P < 2.2 × 10−16. D represents the maximum
absolute difference between the empirical and hypothetical cumulative dis-
The combination of both scores together yielded the highest power to dis-
criminate the two datasets, and as such, the combined value was used to calcu-
late probabilities for our de novo variants to be observed in either database.
Validation experiments. Validation and de novo testing for candidate de novo
mutations was performed using standard Sanger sequencing approaches.
Primers were designed to surround the candidate mutation, and PCR reac-
tions were performed using RedTaq Readymix PCR reaction mix (Sigma-
Aldrich). Primer sequences and PCR conditions are available upon request.
For all de novo mutations identified, an additional control cohort of 75 ethni-
cally matched controls was tested for the presence of the same mutation by
Sanger sequencing. Together with the results from 679 control individuals
from the 1000 Genomes Project as well as the 78 ‘exomes’ present in our in-
house database, the control cohort for the de novo mutations encompassed
1,664 control chromosomes.
26. Marth, G.T. et al. A general approach to single-nucleotide polymorphism discovery.
Nat. Genet. 23, 452–456 (1999).
27. Ng, S.B. et al. Targeted capture and massively parallel sequencing of 12 human
exomes. Nature 461, 272–276 (2009).
28. Pushkarev, D., Neff, N.F. & Quake, S.R. Single-molecule sequencing of an individual
human genome. Nat. Biotechnol. 27, 847–852 (2009).
29. Wang, J. et al. The diploid genome sequence of an Asian individual. Nature 456,
30. Venables, W.N. & Ripley, B.D. Modern Applied Statistics with S (Springer,
4th edn., New York, New York, USA, 2002).
31. Lilliefors, H. On the Kolmogorov–Smirnov test for normality with mean and variance
unknown. J. Am. Stat. Assoc. 62, 399–402 (1967).