SAGE-Hindawi Access to Research
Journal of Nucleic Acids
Volume 2011, Article ID 756905, 6 pages
Validationofa DiagnosticMicroarray forHumanPapillomavirus:
SarahTuttletonArron,1PeterSkewes-Cox,2PhongH. Do,1EricDybbro,1MariaDa Costa,3
Joel M. Palefsky,3and JosephL.DeRisi4,5
1Department of Dermatology, University of California San Francisco, San Francisco, CA 94143, USA
2Biological and Medical Informatics Program, University of California San Francisco, San Francisco, CA 94143, USA
3Department of Medicine, University of California San Francisco, San Francisco, CA 94143, USA
4Department of Biochemistry and Biophysics, University of California San Francisco, San Francisco, CA 94143, USA
5Howard Hughes Medical Institute, Chevy Chase MD 20115-6789, USA
Correspondence should be addressed to Sarah Tuttleton Arron, firstname.lastname@example.org
Received 1 December 2010; Accepted 16 March 2011
Academic Editor: Baochuan Lin
Copyright © 2011 Sarah Tuttleton Arron et al. This is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly
Papillomaviruses have been implicated in a variety of human diseases ranging from common warts to invasive carcinoma of
the anogenital mucosa. Existing assays for genotyping human papillomavirus are restricted to a small number of types. Here,
we present a comprehensive, accurate microarray strategy for detection and genotyping of 102 human papillomavirus types and
validate its use in a panel of 91 anal swabs. This array has equal performance to traditional dot blot analysis with the benefits of
added genotype coverage and the ability to calibrate readout over a range of sensitivity or specificity values.
Papillomaviruses are a group of nonenveloped, epithelio-
tropic DNA viruses that infect the skin and mucous mem-
branes of humans and animals. There are over 100 different
human papillomaviruses (HPV), which are associated with
disease of the skin, mucous membranes, and aerodigestive
tract. HPV infection leads to lesions ranging from common
and genital warts to laryngeal papillomatosis, to epithelial
cancers, particularly cervical carcinoma and a subset of head
and neck squamous cell carcinoma (HNSCC). Detection of
papillomavirus is usually based on molecular assays for viral
nucleic acid .
papillomaviridae. The HPVs fall into five genera: alphapapil-
lomavirus(α-PV), betapapillomavirus(β-PV), gammapapil-
lomavirus (γ-PV), mupapillomavirus (μ-PV), and nupapil-
lomavirus (ν-PV). The mucosal α-PV are the best described,
due to their association with genital malignancy. There are
thirteen types of high-risk mucosal α-PV associated with
cervical and anal intraepithelial neoplasia: 16, 18, 31, 33,
35, 39, 45, 51, 52, 56, 58, 59, and 68. It has recently been
recommended that HPV 66 be added to this high-risk group
. HPV testing is becoming increasingly important as a
screening tool, in combination with cytology, for HPV-
associated neoplasia [3, 4].
Most commercial assays for alphapapillomaviruses are
limited to the high-risk types, which are relevant for use in
the clinical setting but provide limited information in the
research setting. The current FDA-approved tests for HPV
are the DigeneHybrid Capture 2 (HC2)test and the Cervista
HPV16/18 and HR test. HC2 combines antibody capture
and chemiluminescent signal detection to detect 13 high-
risk HPV types (Digene Corporation, Gaithersburg, Md,
USA), while the Cervista HR uses invader chemistry to
detect 14 high risk types (Hologic, Inc., Bedford, Mass,
USA) . While both of these assays have good sensitivity
and specificity for cervical neoplasia, neither specifically
genotypes the HPV infection.
There are currently no genotyping assays that cover mul-
tiple HPV genera. The majority of laboratory tests use
PCR amplification with degenerate or multiplex primer sets
2Journal of Nucleic Acids
followed by genotype readout with sequencing or blot, or
bead-based Multiplexing [6–13]. A number of groups have
developed microarray strategies for detection of HPV, but
these have been limited to targeted PCR and array readout
of a specific subset of HPV types [14–23].
The obstacle to a comprehensive detection strategy is
the sequence diversity of HPVs. Standard PCR assays are
sensitive, butit is difficult to design primers to detect a broad
range ofpapillomavirus types. Here,we present a microarray
strategy for the detection and genotyping of 102 HPV types
representing all genera and validate its use in a large panel of
anal swabs from an HPV cohort study.
2.1. Sample Material. Ninety-one samples were assayed,
drawn from a previously reported cohort study of HPV
infection in the anal canal of homosexual and bisexual
men . The study was approved by the University of
California, San Francisco Committee on Human Research,
and informed consent was obtained from all subjects. As
reported in that study, all samples were typed for HPV with
MY09/MY11 consensus HPV-L1 primers  followed by dot
blot with type-specific biotinylated probes for HPV types 6,
11, 16, 18, 26, 31, 32, 33, 35, 39, 40, 45, 51, 52, 53, 54, 55,
56, 58, 59, 61, 66, 68, 69, 70, 73, 82 variant, 83, and 84,
as well as the following 10 HPV types together in a probe
mixture: 2, 13, 34, 42, 57, 62, 64, 67, 72, and 82 . These
samples were drawn from a population with more multiplic-
ity of types than a typical cervical sample set from healthy
2.2. Array Design. 60nt oligonucleotide probes were de-
genomes listed in GenBank on January 6, 2009 (see Supple-
mentary File 1 in Supplementary Material available online
at doi: 10.406/2011/756905). Genotype-specific probes were
designed using ArrayOligoSelector , a design pack-
age that addresses several design considerations including
uniqueness, sequence complexity, self-annealing as a mea-
sure of secondary structure, and GC content (http://deri-
silab.ucsf.edu/index.php?page=software).Ten probes were
selected for each genome with a target GC content of 30%
and a maximum binding energy cutoff of −45kcal/mol.
Sequences containing AT regions longer than 20bp and less
than 10% of non-AT nucleotides were excluded from probe
selection. A second set of genotype-specific probes were
generatedusing a tiling strategy, with an offset of20nt across
region traditionally used for genotyping. These probes were
their source HPV species and a predicted binding energy
of >−10kcal/mol against all other HPV species. All probes
were filtered against nonspecific binding to the human
Additional probes were designed to cover conserved re-
gions in each genus using a variation on a previously de-
scribed strategy . Each fully sequenced genome was
divided into overlapping 60nt segments offset by 30nt, and
a pairwise nucleotide BLAST  was performed between
each potential probe and each viral genome. The top-
ranked probe was selected, and the process was repeated
iteratively on each sequence lacking significant homology to
the selected probe until each genome was covered by at least
five conserved probes with a predicted binding energy of
<−45kcal/mol. All probes were filtered against nonspecific
binding to the human genome.
Overall, 14,161 60-mer oligonucleotide probes and their
corresponding reverse complements were designed. The
array was synthesized on the Agilent 8x60K SurePrint G3
custom array platform (Agilent Technologies, Santa Clara,
Calif, USA). Each of the 28,322 probes was printed in du-
plicate (Supplementary File 2).
2.3. Array Hybridization. DNA was randomly amplified
and labeled with Cy3 using an adaptation of a previously
described protocol without the reverse transcription step
. In parallel, HeLa cell DNA was used as a positive
amplification and hybridization control, and water was used
as a negative control. Array hybridization was performed
with the AgilentGene Expression Hybridization Kit, adapted
from manufacturer’s specifications with the following mod-
ification: nuclease-free water substituted for fragmentation
buffer. Arrays were washed using Agilent Wash Buffer Kit,
according to manufacturer’s specifications, and scanned
using Agilent Scan Control A.8.1.3 software. Data were
Extraction version 10.5.1.1 Software.
2.4. Array Analysis. Values from replicate spots on the array
were averaged for analysis, and data were analyzed using
the E-Predict algorithm . An E-matrix was generated
from theoretical energy profiles for each of the 102 full-
length HPV genomes in the design set, matched against
each of the array probes. E-Predict compares this theoretical
E-matrix profile to the observed hybridization pattern on
an array to generate a similarity score and rank order of
likelihood for each genome. E-Predict was run with unit
vectorplusquadraticmatrix normalization, unitvectorarray
normalization, and a dot product distance metric.
A set of uninformative oligonucleotides were defined as
probes with significant autofluorescence when hybridized to
waterand probeswith cross-hybridization tohuman DNAin
from E-Predict analysis.
2.5. Sequence Genotyping of HPV. α-HPV PCR was per-
formed using thenested primersetsMY09-MY011 and GP5-
GP6, as previously described in [9, 13]. The PCR products
were visualized by agarose gel electrophoresis, and ampli-
cons of correct size were gel purified using the PureLink
Quick Gel Extraction Kit (Invitrogen, Carlsbad, Calif, USA).
Gel-purified DNA was inserted into pCR2.1 TOPO vec-
tor (Invitrogen) and transformed into chemically competent
TOP10 E. coli (Invitrogen), according to manufacturer’s
specifications. For each DNA sample, 12 transformant
Journal of Nucleic Acids3
Water (n=15) HeLa (n=13)
Area under ROC curve = 0.9795
operating characteristic curve for HPV18.
Area under ROC curve = 0.8287
Figure 2: Receiver operating characteristic curve of the microarray
for 91 anal swab samples tested for 102 HPV genotypes.
colonies were selected for PCR insert amplification using
M13 forward-reverse primers followed by sequencing on the
ABI 3130xl Genetic Analyzer (Applied Biosystems, Carlsbad,
2.6. Statistical Analysis. A receiver-operating characteristic
(ROC) curve was generated to assess the similarity score
for each of the 102 possible HPV genotype outcomes on
array with the actual genotype(s) detected by sequencing
of the 91 samples, for a total of 9282 tests. Sensitivity and
specificity for the array were calculated with similarity score
cutoff values of 0.035 and 0.1. Sensitivity and specificity
were calculated for the blot based on the 29 types covered,
for a total of 2639 tests. Agreement and Cohen’s kappa
coefficients were calculated for the blot versus the array for
the 29 types covered on the blot. All statistical analysis was
performed with Stata 11 (StataCorp LP, College Station, Tex,
28,322 60-mer oligonucleotide probes were designed and
printed on the array. An E-matrix was generated to reflect
the theoretical hybridization energies of each genome in
the design set to each oligonucleotide on the array. This
matrix was used for E-Predict analysis of biological samples
hybridized to the array.
Performance of the array for one representative HPV
type, HPV18, is demonstrated in Figure 1. The HeLa cervical
cancer cell line, which contains integrated HPV18, was
used as a positive control, and water was used as a negative
control. The range of similarity scores for HPV18 showed
a clear demarcation between positive and negative controls
(Figure 1(a)), and a ROC curve was generated with an
area under the curve of 0.9795 (Figure 1(b)), a near-perfect
To characterize the performance of the array across all
genomes, DNA from 91 anal swab samples was randomly
amplified and hybridized to the array. For each specimen
run on the array, a similarity score was generated for each
of the 102 HPV genotypes in the E-matrix, generating 9,282
of PCR and sequencing. A ROC curve was generated with an
area under the curve of 0.8287 (Figure 2).
Based on the ROC curve, two similarity score cut-points
were generated for sensitivity and specificity calculations
(Table 1). A score of 0.035 yielded a sensitivity of 0.72 and
a specificity of 0.76 for the 102 genotypes in the E-matrix. A
score of 0.1 yielded a sensitivity of 0.41 and a specificity of
0.99 for the same 102 types.
Each of the samples had previously been genotyped for
HPV with a PCR and blot assay that covers 29 individual
types and a probe mixture that detects ten additional types
but does not discriminate between them. Compared to PCR
and sequencing, the blot had a sensitivity of 0.41 and a
specificity of 0.98 for the 29 specific types covered, similar
to the array cut-point of 0.1. For the 29 types covered on
4Journal of Nucleic Acids
Table 1: Comparison of array performance to blot. P values: Fisher’s exact test.
Test result +
Test result −
BlotArray (s ≥ 0.1)
Array (s ≥ 0.035)
Table 2: Array detection of 11 HPV types not covered by the blot.
the blot, the two assays had an agreement of 96.6%, kappa
0.6, representing good agreement .
Among the 91 samples, 157 instances of 34 HPV types
were detected by PCR and sequencing. The types detected
were HPV 6, 11, 12, 16, 18, 26, 31, 33, 34, 35, 40, 44, 45,
52, 53, 54, 55, 56, 58, 59, 61, cand62, 66, 67, 70, 71, 72, 81,
82, 83, 84, cand85, 97, and 102. Among those, there were 22
instances of 11 types which were not specifically covered by
the blot strategy. Eighteen of these 22 infections (82%) were
detected on the microarray (Table 2).
We demonstrate the performance of a comprehensive mi-
croarray for detection and genotyping of HPV and validate
this system on a large panel of anal swab samples. Previous
microarrays for HPV genotyping were designed to read out
multiplex or degenerate PCR amplification and targeted
a small number of HPV types. The main benefit of this
system is its accuracy for broad range of HPV genotypes, as
demonstrated by an area under the ROC curve of 0.8287.
Because the nucleic acid amplification is not reliant on
degenerate or multiplex PCR, any HPV type present in the
sample may be detected.
In this sample set, an s value cutoff of 0.1 yielded a
sensitivity of 0.41 and specificity of 0.99. This cutoff favors
specificity over sensitivity, a feature common to many
HPV assays in clinical use. This is based on the prevalent
assumption that extremely low levels of HPV in a sample are
likely not clinically relevant. One example of this in practice
is the blot assay used to characterize this sample set in a
previously reported study . Our array performance was
similar to that of the blot assay (sensitivity 0.41, specificity
0.98) with the benefit of 61 additional HPV types covered on
of 11 HPV types not present on the blot, of which 82% were
detected by the microarray. This demonstrates the benefit of
additional type coverage in a genotyping strategy.
As well as covering more HPV types, our assay is more
flexible than the standard array readout system. Because the
similarity score is a continuous variable, the cutoff can be
tailored to the needs of the assay system. The similarity score
cutoff that optimized both sensitivity and specificity was at
a score of 0.035 (Figure 2). This yielded a sensitivity of 0.71
and specificity of 0.76 compared with PCR and sequencing.
We validated the performance of this microarray by a
comparison to a gold standard of sequence genotyping.
No previous study has reported microarray sensitivity and
specificity with thistypeofvalidation.Some studiesreported
limits of detection and the absence of cross-hybridization
but did not address true sensitivity and specificity [16, 20,
21]. Other studies compared array performance to other
targeted HPV assays rather than to a gold standard of
sequencing. Thus, the authors were only able to report
interassay agreement or Cohen’s kappa. Ermel et al. com-
pared performance of α-PV detection with the Digene
Hybrid Capture II Assay, the Roche Linear Array Assay,
and the Kurabo GeneSquare Assay and reported a sample
proportion of agreement ranging from 0.86 to 0.89 between
the three assays. In two studies assessing multiplex PCR and
array primer extension (APEX), Gheit et al. compared their
arrays to reverse line blot or to one-step PCR and reverse
hybridization with overall kappas of 0.55 and 0.5. In those
studies, only a few samples were selected for sequencing to
confirm array detection of types not seen on the other assays
[14, 15]. In our study, we measured a sample proportion
of agreement of 0.97 between the array and blot, with a
kappa of 0.6. This level of agreement demonstrates the
excellent performance of our system compared to previ-
ously reported arrays.The drawback to comparing assays
with different types of coverage is that the agreement can
only be assessed for the overlapping types, which under-
estimates the utility of a comprehensive array such as this
A number of arrays targeting the high-risk α-PV have
been evaluated by calculating the sensitivity and specificity
of the test for a cytologic diagnosis of neoplasia, which may
Journal of Nucleic Acids5
be clinically relevant for a screening diagnostic but does not
address the performance of the assay in genotyping HPV
[18, 19, 22,23,31].Futurestudiesmay address thesensitivity
and specificity of our assay for cervical or anal intraepithelial
The theoreticalE-matrix ofenergyprofilesusedtogener-
ate similarity scores was created from the 102 HPV genomes
used for oligonucleotide probe design. However, the E-
genomes if desired, allowing for an updated genotyping
strategy without the synthesis of additional probes. This
flexibility will lend itself to future studies as the number of
complete papillomavirus sequences continues to expand. As
see a shift in HPV prevalence and the rise of less common
types. This assay is well suited to detect these trends.
Overall, this microarray provides a comprehensive and
spectrum ofHPVtypes. It will lenditself to high-throughput
applications such as screening of tumor cohorts and large
S. T. Arron is supported by NIH/NCRR/OD UCSF-CTSI
Grant no. KL2 RR024130. P. Skewes-Cox is supported by
a PhRMA Foundation Predoctoral Informatics Fellowship.
J. L. DeRisi is supported by the Howard Hughes Medical
Institute. The authors wish to thank W. John Boscardin and
WIP-B for advice on statistical analysis. S. T. Arron and P.
Skewes-Cox contributed equally to this work.
 Fields Virology, Lippincott Williams & Wilkins, Philadelphia,
Pa, USA, 2006.
 V. Cogliano, R. Baan, K. Straif, Y. Grosse, B. Secretan, and
F. El Ghissassi, “Carcinogenicity of human papillomaviruses,”
Lancet Oncology, vol. 6, no. 4, p. 204, 2005.
 M.H.Mayrand,E.Duarte-Franco, I.Rodrigues etal.,“Human
papillomavirus DNA versus Papanicolaou screening tests for
cervical cancer,”NewEnglandJournal of Medicine,vol.357,no.
16, pp. 1579–1588, 2007.
 G. Ronco, P. Giorgi-Rossi, F. Carozzi et al., “Efficacy of human
papillomavirus testing for the detection of invasive cervical
cancers and cervical intraepithelial neoplasia: a randomised
controlled trial,” The Lancet Oncology, vol. 11, no. 3, pp. 249–
 L. R. Johnson, C. R. Starkey, J. Palmer et al., “A comparison
of two methods to determine the presence of high-risk HPV
cervical infections,” American Journal of Clinical Pathology,
vol. 130, no. 3, pp. 401–408, 2008.
 O. Forslund, A. Antonsson, P. Nordin, BO. Stenquist, and B.
G. Hansson, “A broad range of human papillomavirus types
detected with a general PCR method suitable for analysis
of cutaneous tumours and normal skin,” Journal of General
Virology, vol. 80, no. 9, pp. 2437–2443, 1999.
 O. Forslund, H. Ly, and G. Higgins, “Improved detection of
cutaneous human papillomavirus DNA by single tube nested
’hanging droplet’ PCR,” Journal of Virological Methods, vol.
110, no. 2, pp. 129–136, 2003.
 Y. J.Lei,C. Gao,R. Anetal.,“DevelopmentofamultiplexPCR
method for detecting and typing human papillomaviruses in
verrucae vulgaris,” Journal of Virological Methods,vol. 147, no.
1, pp. 72–77, 2008.
 M. Manos, Y. Ting, D. Wright, A. Lewis, T. Broker, and S.
Wolinsky, “Use of polymerase chain reaction amplification
for the detection of genital human papillomaviruses,” in
Molecular Diagnostics of Human Cancer, vol. 7 of Cancer Cells,
pp. 209–214, Cold Spring Harbor Laboratory, 1989.
 I. Nindl, A. K¨ ohler, M. Gottschling et al., “Extension of the
typing in a general-primer-PCR reverse-line-blotting system
to detect all 25 cutaneous beta human papillomaviruses,”
Journal of Virological Methods, vol. 146, no. 1-2, pp. 1–4, 2007.
 I. Sabol, M. Salakova, J. Smahelova et al., “Evaluation of dif-
ferent techniques for identification of human papillomavirus
types of low prevalence,” Journal of Clinical Microbiology, vol.
46, no. 5, pp. 1606–1613, 2008.
 M. Schmitt, I. G. Bravo, P. J. F. Snijders, L. Gissmann, M.
Pawlita, and T. Waterboer, “Bead-based multiplex genotyping
of human papillomaviruses,” Journal of Clinical Microbiology,
vol. 44, no. 2, pp. 504–512, 2006.
 P. J.Snijders, A. J. van den Brule, M. V. Jacobs,R. P. Pol, andC.
J.Meijer, “HPVDNAdetection andtypingincervicalscrapes,”
Methods in Molecular Medicine, vol. 119, pp. 101–114, 2005.
 T. Gheit, G. Billoud, M. N. C. De Koning et al., “Development
of a sensitive and specific multiplex PCR method combined
with DNA microarray primer extension to detect betapapillo-
mavirus types,” Journal of Clinical Microbiology, vol. 45, no. 8,
pp. 2537–2544, 2007.
 T. Gheit, S. Landi, F. Gemignani et al., “Development of a
sensitive and specific assay combining multiplex PCR and
humanpapillomavirustypes,” Journalof Clinical Microbiology,
vol. 44, no. 6, pp. 2025–2031, 2006.
 V. Albrecht, A. Chevallier, V. Magnone et al., “Easy and fast
detection and genotyping of high-risk human papillomavirus
by dedicated DNA microarrays,” Journal of Virological Meth-
ods, vol. 137, no. 2, pp. 236–244, 2006.
 A. Ermel, B. Qadadri, A. Morishita et al., “Human papillo-
mavirus detection and typing in thin prep cervical cytologic
specimens comparingtheDigene Hybrid Capture II Assay, the
Roche Linear Array HPV Genotyping Assay, and the Kurabo
GeneSquare Microarray Assay,”Journal of Virological Methods,
vol. 169, no. 1, pp. 154–161, 2010.
 C. J. Kim, J. K. Jeong, M. Park et al., “HPV oligonucleotide
microarray-based detection of HPV genotypes in cervical
neoplastic lesions,” Gynecologic Oncology, vol. 89, no. 2, pp.
 G. J. Nuovo, D. Bartholomew, W. W. Jung et al., “Correlation
of pap smear, cervical biopsy, and clinical follow-up with
an HPV typing microarray system,” Diagnostic Molecular
Pathology, vol. 17, no. 2, pp. 107–111, 2008.
 T. Oh, C. Kim, S. Woo et al., “Development and clinical eval-
uation of a highly sensitive DNA microarray for detection and
genotyping of human papillomaviruses,” Journal of Clinical
Microbiology, vol. 42, no. 7, pp. 3272–3280, 2004.
 S. Ramamoorthy, Y. -T. Liu, L. Luo, K. Miyai, Q. Lu, and J.
M. Carethers, “Detection of multiple human papillomavirus
genotypes in anal carcinoma,” Infectious Agents and Cancer,
vol. 5, no. 1, article 17, 2010.
 H. J. An, N. H. Cho, S. Y. Lee et al., “Correlation of cervical
carcinoma and precancerous lesions with human papillo-
mavirus (HPV) genotypes detected with the HPV DNA chip
6Journal of Nucleic Acids Download full-text
microarray method,” Cancer, vol. 97, no. 7, pp. 1672–1680,
 T. S. Hwang, J. K. Jeong, M. Park, H. S. Han, H. K. Choi,
and T. S. Park, “Detection and typing of HPV genotypes in
various cervical lesions by HPV oligonucleotide microarray,”
Gynecologic Oncology, vol. 90, no. 1, pp. 51–56, 2003.
 J. M. Palefsky, E. A. Holly, M. L. Ralston, and N. Jay,
“Prevalence and risk factors for human papillomavirus infec-
tion of the anal canal in human immunodeficiency virus
(HIV)-positive and HIV-negative homosexual men,” Journal
of Infectious Diseases, vol. 177, no. 2, pp. 361–367, 1998.
 Z. Bozdech, J. Zhu, M. P. Joachimiak, F. E. Cohen, B. Pulliam,
and J. L. DeRisi, “Expression profiling of the schizont and
trophozoite stages of Plasmodium falciparum with a long-
oligonucleotide microarray,” Genome Biology, vol. 4, no. 2,
article R9, 2003.
 D. Wang, L. Coscoy, M. Zylberberg et al., “Microarray-based
detection and genotyping of viral pathogens,” Proceedings
of the National Academy of Sciences of the United States of
America, vol. 99, no. 24, pp. 15687–15692, 2002.
 S. F. Altschul, W. Gish, W. Miller, E. W. Myers, and D.
J. Lipman, “Basic local alignment search tool,” Journal of
Molecular Biology, vol. 215, no. 3, pp. 403–410, 1990.
 D. Wang, A. Urisman, YU. T. Liu et al., “Viral discovery and
sequence recovery using DNA microarrays,”PLoS Biology, vol.
1, no. 2, article e2, 2003.
 A. Urisman, K. F. Fischer, C. Y. Chiu et al., “E-Predict: a
computational strategy for species identification based on
observed DNA microarray hybridization patterns,” Genome
Biology, vol. 6, no. 9, article R78, 2005.
 D. G. Altman, Practical Statistics for Medical Research, Chap-
man & Hall, London,UK, 1991.
 N. H. Cho, H. J. An, J. K. Jeong et al., “Genotyping of 22
human papillomavirus types by DNA chip in Korean women:
comparison with cytologic diagnosis,” American Journal of
Obstetrics and Gynecology, vol. 188, no. 1, pp. 56–62, 2003.