Published Ahead of Print on January 23, 2012 as 10.1200/JCO.2011.39.

2316
The latest version is at http://jco.ascopubs.org/cgi/doi/10.1200/JCO.2011.39.2316

JOURNAL OF CLINICAL ONCOLOGY S P E C I A L A R T I C L E

Cancer Genomics: Technology, Discovery, and Translation

Ben Tran, Janet E. Dancey, Suzanne Kamel-Reid, John D. McPherson, Philippe L. Bedard,
Andrew M.K. Brown, Tong Zhang, Patricia Shaw, Nicole Onetto, Lincoln Stein, Thomas J. Hudson,
Benjamin G. Neel, and Lillian L. Siu
Ben Tran, Philippe L. Bedard, and Lillian
L. Siu, Princess Margaret Hospital, A B S T R A C T
University Health Network, University
of Toronto; Janet E. Dancey, John D. In recent years, the increasing awareness that somatic mutations and other genetic aberrations drive
McPherson, Andrew M.K. Brown, human malignancies has led us within reach of personalized cancer medicine (PCM). The implemen-
Nicole Onetto, Lincoln Stein, and tation of PCM is based on the following premises: genetic aberrations exist in human malignancies; a
Thomas J. Hudson, Ontario Institute for subset of these aberrations drive oncogenesis and tumor biology; these aberrations are actionable
Cancer Research; Suzanne Kamel-Reid,
(defined as having the potential to affect management recommendations based on diagnostic,
Tong Zhang, and Patricia Shaw, Toronto
General Hospital, University Health
prognostic, and/or predictive implications); and there are highly specific anticancer agents available that
Network, University of Toronto; John D. effectively modulate these targets. This article highlights the technology underlying cancer genomics
McPherson, Nicole Onetto, Lincoln and examines the early results of genome sequencing and the challenges met in the discovery of new
Stein, Thomas J. Hudson, and Benja- genetic aberrations. Finally, drawing from experiences gained in a feasibility study of somatic mutation
min G. Neel, University of Toronto; and genotyping and targeted exome sequencing led by Princess Margaret Hospital–University Health
Benjamin G. Neel, Campbell Family
Network and the Ontario Institute for Cancer Research, the processes, challenges, and issues involved
Cancer Research Institute, Ontario
Cancer Institute, University Health
in the translation of cancer genomics to the clinic are discussed.
Network, University of Toronto,
Toronto, Canada. J Clin Oncol 30. © 2012 by American Society of Clinical Oncology
Submitted August 29, 2011; accepted
November 16, 2011; published online Since the Human Genome Project, the emerg-
ahead of print at www.jco.org on INTRODUCTION
January 23, 2012.
ing scientific era of “omics” has revolutionized the
In recent years, an increasing appreciation and iden- study of cancer. Although cancer is recognized as a
Terms in blue are defined in the glos-
sary, found at the end of this article tification of somatic mutations and other genetic disease driven fundamentally by genetic changes,
and online at www.jco.org. aberrations that drive human malignancies have led the somatic events that drive the multistep progres-
Authors’ disclosures of potential con- us within reach of personalized cancer medicine sion of carcinogenesis are not well understood, even
flicts of interest and author contribu- (PCM). The US National Cancer Institute defines in the most studied cancer types.3 The International
tions are found at the end of this
article.
personalized medicine as “a form of medicine that Cancer Genome Consortium (ICGC) is coordinat-
uses information about a person’s genes, proteins, ing efforts aimed at identifying all genomic altera-
Corresponding author: Lillian L. Siu,
MD, FRCPC, Princess Margaret Hospi- and environment to prevent, diagnose, and treat tions significantly associated with cancer, including
tal, Drug Development Program, 610 disease.”1 The implementation of PCM is based on genomic loss or amplification, mutations in coding
University Ave, Ste 5-718, Toronto,
the following premises: genetic aberrations exist in regions, chromosomal rearrangements, aberrant
Ontario, M5G 2M9, Canada; e-mail:
[email protected].
human malignancies; a subset of these aberrations methylation, and expression profiles. Through the
© 2012 by American Society of Clinical
drives oncogenesis and tumor biology; these aberra- ICGC, the discovery pillar targets the decoding of
Oncology tions are actionable (have potential to affect man- cancer genomes. Much of the discovery is depen-
0732-183X/12/3099-1/$20.00
agement recommendations based on diagnostic, dent on advances in molecular diagnostics, particu-
prognostic, and/or predictive implications); and
DOI: 10.1200/JCO.2011.39.2316 larly genome sequencing, within the technology
highly specific anticancer agents are available that
pillar. The improved timeliness and cost associated
effectively modulate these targets. The National
with genome sequencing have driven discovery not
Cancer Institute Office of Cancer Genomics, estab-
only in cancer genomics but also in the final pillar of
lished to facilitate PCM through validation of these
clinical translation. Until now, the majority of the
key premises, articulates the following mission and
focus within cancer genomics lay in discovery. As
goals2: First, enhance the understanding of the mo-
lecular mechanisms of cancer; second, accelerate our molecular understanding of cancer improves,
genomic science and technology development; and the prospect of applying genomic knowledge in the
third, translate genomic data to improve cancer pre- clinic becomes increasingly tangible. However, spe-
vention, early detection, diagnosis, and treatment. cific challenges in the scientific, regulatory, and eth-
Hence, the focus of this article will revolve around ical domains remain to be overcome before PCM
the three pillars that support cancer genomics: dis- can become a reality. This article highlights the tech-
covery, technology, and translation. nology that empowers cancer genomics, examines

© 2012 by American Society of Clinical Oncology 1

Downloaded from jco.ascopubs.org on February 20, 2013. For personal use only. No other uses without permission.
Copyright © 2012 American Society of Clinical Oncology. All rights reserved.
Copyright 2012 by American Society of Clinical Oncology
 Tran et al

the early results of genome sequencing and the challenges met in the biologic processes that results in longer read lengths, averaging 964
discovery of new genetic aberrations, and discusses the processes in- bases in published articles,4,11 and more than 2,000 bases in more
volved in the translation of cancer genomics to the clinic. recent applications at our institute. Instead of relying on amplified
DNA, single-molecule sequencing detects the specific sequence of
each individual DNA strand. Ion Torrent PGM uses nonoptical DNA
TECHNOLOGY sequencing. Rather than detecting nucleotide incorporation optically,
PGM uses a semiconductor that senses the ions produced as nucleo-
The field of cancer genomics is growing rapidly as a result of revolu- tides are incorporated. Although read length currently averages fewer
tionary advances in DNA sequencing technologies. In this section, we than 200 bases, accuracy is high, and run time is short, potentially
review the technologic developments that have catalyzed increased allowing for real-time clinical application.10,12
understanding of cancer biology: whole genome sequencing (WGS), The first human genome sequence cost more than $2 billion and
targeted sequencing, genotyping, and bioinformatics. took a decade to complete.20 However, advances in cost and timeliness
gained through these novel platforms bring the $1,000 genome target
WGS of the National Institutes of Health within reach. As genome sequenc-
WGS is the backbone technology that supports the in-depth ing becomes more affordable and accessible, our understanding of the
sequencing of cancer genomes. Genome sequencing consists of three molecular basis of cancer is expected to improve exponentially.
phases: sample preparation, physical sequencing, and reconstruction.4
In sample preparation, the target genome is broken into fragments.4 Targeted Genome Sequencing
During physical sequencing, individual bases in each fragment are Although cheaper than Sanger sequencing, WGS remains expen-
identified in order; the number of individual bases identified contig- sive on a grand scale, with current costs of $10,000 to $35,0007 per
uously is defined as the read length.4 During reconstruction, bioinfor- human genome, exclusive of labor and other expenses. Targeted se-
matics software aligns overlapping reads from each fragment, allowing quencing refers to strategies that enrich the input for DNA regions of
the original genome to be constructed; the longer the read length, the interest,7 such as the whole exome or the cancer genome (ie, genes
easier the reconstruction.5 Traditionally, genome sequencing has been potentially involved in tumor biology). Many of the platforms used for
costly and time consuming. However, new technology has diminished WGS also are used for targeted sequencing, although polymerase
both of these impediments.4,5 chain reaction (PCR) amplification of targeted regions or hybridiza-
First-generation sequencing, or Sanger sequencing, has been tion of the test DNA to specific arrays of oligonucleotides correspond-
the workhorse of DNA sequencing for almost 30 years.4 Although ing to the desired target sequences is required. In addition to reducing
significant improvements in optimization, miniaturization, multi- the cost per sample, these approaches increase coverage of areas of
plexing, automation, and pipeline integration have occurred, the interest, which may overcome problems of cancer cell cellularity in
fundamental technology has not changed significantly.4 Sanger tumor specimens and increase accuracy.7 As WGS costs remain high,
sequencing can produce read lengths of up to 1,000 bases, consid- targeted sequencing, particularly exome sequencing, is likely to dom-
erably longer than second-generation sequencing platforms.5 It is inate near-term sequencing strategies.21
an effective method, the long read lengths and high accuracy of
which have resulted in monumental accomplishments, including Cancer Genotyping
completion of the Human Genome Project.4,6 However, limita- The increasing number of targeted therapeutics, the antitumor
tions of high cost and low throughput (small amount of data gener- activity of which is based on the presence of specific biomarkers, has
ated per unit of time) have led to the development of next-generation created a growing need for real-time detection of recognized genetic
sequencing (NGS).4,5,7 aberrations in clinical samples in a cost-effective and timely manner.
NGS platforms consist of second- and third-generation technol- Given the observation that some cancer mutations occur at similar
ogies, described in depth by Metzker.4 Both are more economical than DNA bases in tumors from different patients (so-called recurrent
Sanger and have higher throughput. Second-generation platforms are mutations), it is possible to use assays that test for single bases (a
dominated by cyclic array– or flush and scan– based sequencing. process referred to as mutation genotyping). Given that the number of
Strands of fragmented DNA are amplified, and bases are then added clinically validated recurrent and predictive mutations are few, meth-
sequentially using DNA polymerase. Excess reagent is washed out, ods such as PCR-based restriction fragment length polymorphism are
imaging then identifies the base incorporated, and the process is re- currently being used for somatic mutation genotyping in individual
peated.8 This repetitive process leads to millions of reads, each of patients with cancer.13 However, the repertoire of recurrent mutations
limited length (approximately 50 to 400 bases), creating a challenge for is increasing, and there is interest in testing these to evaluate their role
genome reconstruction.4,5,7,9 Several second-generation platforms are as predictive mutations for the numerous molecularly targeted agents
commercially available, each with distinct differences (Table 1). Al- in development. This leads to a need for higher-throughput genotyp-
though NGS has improved cost and throughput, its disadvantages ing methods.
include short read length, complex sample preparation, need for am- High-throughput genotyping platforms, consisting of multi-
plification, long time to results, and significant data storage and inter- plexed assays and microarrays, have been successfully used for
pretation requirements.5,11 genotyping clinical samples.15,22-24 Table 1 details several of these
Third-generation sequencing technologies include novel plat- platforms, including the Taqman OpenArray Genotyping system (us-
forms, such as the PacBio RS (Pacific Biosciences, Menlo Park, CA) ing Taqman genotyping assays; Applied Biosystems, Carlsbad, CA),
and Ion Torrent PGM (Life Technologies, Carlsbad, CA; Table 1). ABI 3730 DNA Analyzer (using SNaPshot assays; Applied Biosys-
PacBio RS uses a process called single-molecule, real-time detection of tems), iScan platform (using Goldengate assays; Illumina, San Diego,

2 © 2012 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Downloaded from jco.ascopubs.org on February 20, 2013. For personal use only. No other uses without permission.
Copyright © 2012 American Society of Clinical Oncology. All rights reserved.
 Cancer Genomics

Table 1. Sequencing and Genotyping Platforms

Platform Method Application Comment
Sequencing
First generation (Sanger sequencing)
Sanger4 Strands of fragmented DNA are resolved on Targeted sequencing; whole Despite high accuracy and successes such
gel and distributed in order of length, genome sequencing; as the first human genome, several
with end base labelled genotyping limitations, particularly low throughput,
have led to increased use of NGS
technologies
Second generation (cyclic Stands of fragmented DNA are amplified; Targeted sequencing; whole Higher throughput has provided significant
array–based sequencing) then bases are added sequentially using genome sequencing advantages; however, limitations such as
DNA polymerase; excess reagent is sample preparation, short read lengths,
washed out, imaging identifies base and relatively slow run time have limited
incorporated, and process repeats clinical use; newer versions (such as
MiSeq 关Illumina兴or 454 Junior 关Roche兴)
sacrifice genome coverage for faster run
time to become more amenable to
clinical application
454 (Roche, Basel, Pyrophosphate released at time of base
Switzerland)4,5,10 incorporation
HiSeq (Illumina, San Diego, Fluorescent-labelled nucleotides added
4,5,9
CA) simultaneously
SOLiD 4 (Life Technologies, Driven by DNA ligase instead of DNA
Carlsbad, CA)4,5,10 polymerase
Third generation (novel technologies)
PacBio RS (Pacific Biosciences, Single-molecule real-time sequencing; Targeted sequencing; whole Results in long read lengths, short run time,
Menlo Park, CA)4,10,11 imaging of dye-labelled nucleotides as genome sequencing and high throughput with simple sample
they are incorporated during DNA preparation; potential for clinical
synthesis by single DNA polymerase application
molecule
Ion Torrent PGM (Life Nonoptical DNA sequencing; massively Targeted sequencing; whole Low technology cost and short run time;
Technologies)10,12 parallel semiconductor senses ions genome sequencing potential for clinical application
produced as nucleotides are incorporated
by DNA polymerase-based synthesis
Genotyping
Restricted fragment length Uses restriction enzymes to fragment DNA Single somatic mutation Allows detection of low-frequency
polymorphism13,14 in presence of targeted mutation; then analysis mutations (⬎ 4%) but has low
gel electrophoresis separates resulting throughput and is dependent upon
fragments, identifying mutation subjective visual interpretation; still used
in some centers for KRAS mutation
testing; however, not feasible method for
high-throughput genotyping
Taqman OpenArray Genotyping Uses allele-specific PCR and dye-labelled Somatic mutation analysis; Effective and accurate high-throughput
System (Applied Biosystems, probes (Taqman assay) combined with SNP genotyping genotyping platform
Carlsbad, CA)14,15,16 fluorescent readout systems
MassARRAY (Sequenom, San Diego, Uses allele-specific PCR combined with Somatic mutation analysis; Effective and accurate high-throughput
CA)14,15,16 MALDI-TOF mass spectrometry to detect SNP genotyping; gene genotyping platform; able to detect low-
mutations/SNPs expression analysis; frequency mutations (⬎ 10%); premade
methylation analysis (Oncocarta) and customized mutation
panels available
ABI PRISM 3100 Genetic Analyzer Uses allele-specific PCR with Somatic mutation analysis; Effective and accurate high-throughput
(Applied Biosystems)14,17,18 oligonucleotide primers and labelled SNP genotyping; gene genotyping platform
nucleotides for primer extension expression analysis;
(SNaPshot assay) combined with capillary methylation analysis
electrophoresis and optical imaging
15,16
iScan (Illumina) Uses allele- and locus-specific PCR with Somatic mutation analysis; Effective and accurate high-throughput
oligonucleotide primers; hybridization of SNP genotyping; gene genotyping platform
assay products onto BeadChip; then expression analysis;
imaging of fluorescent signals methylation analysis
Gene Titan (Affymetrix, Santa Clara, Uses microarray technology and GeneChip Somatic mutation analysis; Effective and accurate high-throughput
CA) arrays SNP genotyping genotyping platform
aCGH platform (Agilent, Santa Clara, Uses microarray technology and CGH arrays aCGH Effective platform for analysis of copy
CA)19 (including Agilent and Oxford Gene number variations with high resolution
Technology 关Oxford, United Kingdom兴 and high throughput
arrays) to detect copy number variations

Abbreviations: aCGH, array-based CGH; CGH, comparative genomic hybridization; MALDI-TOF, matrix-assisted laser desorption/ionization–time of flight; NGS,
next-generation sequencing; PCR, polymerase chain reaction; SNP, single nucleotide polymorphism.

www.jco.org © 2012 by American Society of Clinical Oncology 3

Downloaded from jco.ascopubs.org on February 20, 2013. For personal use only. No other uses without permission.
Copyright © 2012 American Society of Clinical Oncology. All rights reserved.
 Tran et al

CA), Affymetrix genotyping arrays (Santa Clara, CA), and MassARRAY

Table 2. Role of Bioinformatics in Cancer Genomics
platform (using mass spectrometry with matrix-assisted laser desorption/
ionizationtime-of-flightanalysis;Sequenom,SanDiego, CA).14-18 These Bioinformatics
Function Background
high-throughput genotyping platforms can analyze hundreds to mil-
Genome alignment Alignment and reconstruction requires reads
lions of germline and/or somatic variants simultaneously and are and sufficiently long enough to be mapped accurately
distinctly different from sequencing technologies. Whereas DNA se- reconstruction onto reference genome sequence13,15,21
quencing can detect any sequence variant in the gene(s) evaluated, Mapping processes must efficiently handle millions of
generated sequences while being robust in
genotyping detects only known variants that have been selected for presence of sequencing errors and SNPs13,21
analysis. Because multiplexed assays and microarrays are relatively Existing sequencing alignment tools like BLAST or
inexpensive and provide results rapidly, they are currently the most BLAT are adequate for long reads produced by
Sanger sequencing, but for short reads produced by
common technologies used for both somatic and germline mutation NGS, newer alignment tools are being developed to
genotyping in clinical samples. allow for mismatches and/or gaps13,21
Simple somatic mutations are only one of several types of SNPs identified through this process must be carefully
analysed to ensure they are real and not a result of
genetic aberrations that have the potential to be predictive bio- technology-specific errors13,21
markers. Translocations, DNA amplifications, deletions, methyl- Base calling Base calling is process of converting sequencing
ation, and gene expression also are important, and assaying for signals into base and is essential for SNP and
somatic variant identification31
these aberrations also can be performed. In the clinic today, fluo- Improvements in base calling accuracy are essential to
rescence in situ hybridization (FISH) is the gold standard for iden- reduce false positives and lead to more reliable
tification of ERBB2 amplification.25-27 However, like restriction identification of germline and somatic variants31,33
De novo genome De novo genome assembly is like solving a large
fragment length polymorphism, FISH has a low throughput. To iden- assembly jigsaw puzzle without knowing the final picture
tify and validate gene copy number changes that include deletions, Several assembly tools have been adapted or
gains, and amplifications (when gene copy number is greater than 10), independently developed for generating assemblies
from short reads30,31
a higher-throughput technology is required. Array-based comparative Genome browsing Genome browsing and annotation enable millions of
genomic hybridization is a molecular-cytogenetic method for detec- and annotation sequences to be available to biomedical community
tion of gene copy numbers that has high resolution and high through- through easily accessible and user-friendly systems;
essential for collaboration and progress in research
put. Technologies such as the Agilent array-based comparative Commonly used browsers include EntrezGene
genomic hybridization platform (Santa Clara, CA) are able to perform browser, University of California Santa Cruz genome
high-throughput genotyping for gene copy number variations in clin- browser, and European Bioinformatics
Institute/Ensemble browser31
ical samples.19,28 Commonly used browsers containing cancer mutation
Multiplexed assays and microarrays are currently the dominant datasets include COSMIC and ICGC
technologies in high-throughput genotyping of somatic mutations, National Centre for Biotechnology Information SNP
database stores millions of SNPs,31 which can be
gene copy number variations, and other alterations affecting gene useful in classifying mutations into known germline
expression or DNA methylation.29 However, as costs associated with variants
targeted or whole-genome sequencing fall, NGS platforms may be- Abbreviations: BLAST, Basic Local Alignment Search Tool; BLAT, BLAST-Like
come the preferred option. Alignment Tool; COSMIC, Catalogue of Somatic Mutations in Cancer; ICGC,
International Cancer Genome Consortium; NGS, next-generation sequencing;
SNP, single nucleotide polymorphism.
Bioinformatics
Bioinformatics is the application of statistics and computer
science to biology. It includes both information management (eg,
genomic databases and visualization) and algorithm development, sive catalogues of genomic abnormalities in 500 tumors from each of
particularly for the assembly, annotation, and comparison of 50 different cancer types” and “to accelerate research into the causes
genomes.30-33 As detailed in Table 2, these bioinformatic functions are and control of cancer.”43
an essential component of cancer genomics. A detailed discussion of Whereas ICGC projects aim to develop a molecular map of the
bioinformatics in the age of NGS is beyond the scope of this review; genetic aberrations involved in cancer, genome-wide association stud-
instead, the reader is referred to an excellent review of this topic by Pop ies (GWAS) investigate the inherited basis of cancer by comparing
et al.30 common DNA variations in a large set of unrelated patient cases and
controls and identifying genetic variants associated with altered risk.44
Although GWAS are important in the larger scheme of cancer genom-
DISCOVERY ics, this article focuses predominantly on acquired genetic aberrations
that arise in the genomes of cancer cells.
The discovery of genetic aberrations in human cancers has identified Results from ICGC studies of glioblastoma multiforme (GBM),
potential therapeutic targets and provided key insights into the mech- ovarian carcinoma, and chronic lymphocytic leukemia have been
anisms underlying tumorigenesis,34-41 as described in an excellent published.38,45,46 These studies surveyed for gene mutations, DNA
review by Stratton et al.42 The ICGC, formed in 2008 and incorporat- copy number, gene expression, and methylation in large cohorts. The
ing the Cancer Genome Project of the United Kingdom and the GBM study identified a substantial proportion of tumors with MGMT
Cancer Genome Atlas of the United States, coordinates research proj- promoter methylation, now known to be a predictive biomarker for
ects that aim to comprehensively elucidate genomic changes present temozolomide sensitivity.38 The ovarian study identified impaired
in multiple cancers.43 Its primary goals are to “generate comprehen- homologous recombination in approximately 50% of tumors, a

4 © 2012 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Downloaded from jco.ascopubs.org on February 20, 2013. For personal use only. No other uses without permission.
Copyright © 2012 American Society of Clinical Oncology. All rights reserved.
 Cancer Genomics

potential predictor for benefit from poly (ADP-ribose) polymerase hepatocellular carcinoma, and multiple myeloma have also been
inhibitors.45 The chronic lymphocytic leukemia study demonstrated sequenced. Results from each study illustrate the effectiveness of
that NOTCH1 and MYD88 mutations are associated with distinct cancer genome sequencing in furthering our understanding
clinical subgroups with specific biologic features.46 These results vali- of cancer.35-37,40,47-49
date the role of cancer genomics in achieving PCM.
Epigenetics and Gene Expression
Sequenced Cancer Genomes Epigenetic aberrations in cancer, such as global hypomethylation
Much of the focus in cancer genomics has centered on sequenc- of DNA, hypermethylation of tumor-suppressor genes, and inactiva-
ing cancer genomes and identifying potential driver mutations (ie, tion of microRNA by DNA methylation, are being systematically
mutations that are functionally critical in the tumor). Cancer genomes studied, because it is clear they can have a significant impact on tumor
from several tumor types have been sequenced and published (Table biology and treatment outcomes.50 Using transcriptional profiling,
3). The earliest comprehensive survey of human genes in cancer by gene expression signatures are also being studied extensively in can-
sequencing methods examined the whole exome (20,661 genes) of 22 cer.51 However, because of challenges associated with reproducibility,
human GBM samples, correlating sequencing results with patient only a few validated discoveries have been made to date. These include
outcome data. This study resulted in the unexpected but important the novel molecular classification of breast cancer initially reported by
finding of IDH1 mutations as a potential favorable prognostic bio- Perou et al52 and the development of validated recurrence scores for
marker.39 These initial results provided early validation of the utility of early breast cancer such as OncotypeDX (Genomic Health, Redwood
genome sequencing in cancer. City, CA) and MammaPrint (Agendia, Irvine, CA).53,54 Gene expres-
Whole genomes (or exomes) of acute myeloid leukemia, mel- sion signatures can contribute to PCM by being diagnostic, prognos-
anoma, small-cell lung cancer, prostate cancer, pancreatic cancer, tic, and/or predictive, and such profiling efforts can be broadened to

Table 3. Sequenced Cancer Genomes

Novel
No. of Genome Mutations in
Author Tumor Samples Tissue Type or Exome Novel Mutations Coding Regions Comment
Ding et al34 Basal-like breast 1 Blood, primary, Genome 27,173, primary; 200, primary; 48 validated somatic mutations
cancer metastasis, 51,710, 225, present in all three tumor
xenograft metastasis; metastasis; tissues, with two additional
109,078, 328, mutations in metastasis
xenograft xenograft
Mardis et al35 AML 1 Tumor, skin Genome 20,256 113 Recurrent mutations in IDH1
discovered
Ley et al41 AML 1 Tumor, skin Genome 31,632 241 Eight newly defined somatic
mutations for AML
Pleasance et al37 Malignant 1 Cell line, Genome 33,345 292 Identification of mutation
melanoma lymphoblastoid signature caused by
cell line exposure to ultraviolet light
Pleasance et al36 Small-cell lung 1 Cell line, Genome 22,190 134 Identification of mutation
cancer lymphoblastoid signature caused by
cell line exposure to tobacco smoke
Parsons et al39 GBM 22 Seven tumors; 15 Exomeⴱ NA 47 (mean) Recurrent mutations in IDH1
xenograft, blood discovered
Berger et al40 Prostate cancer 7 Tumor, blood Genome 3,866 (median) 20 (median) Four of seven patients
harbored events disrupting
PI3K pathway
Jones et al47 Pancreatic cancer 24 Tumor, normal Exomeⴱ NA 63 (mean) Identified 12 pathways,
duodenum component genes of which
were most altered in
pancreatic cancer
Chapman et al48 Multiple myeloma 23† Bone marrow, blood Genome 7,450 (mean) 35 (mean) BRAF mutations identified in
4% of samples
Totoki et al49 HCC 1 Tumor, blood Genome 12,401 88 Significant intratumoral
heterogeneity demonstrated
by TSC1 mutation frequency
of 13%; only detected by
whole exome sequencing at
higher sequence depth
Puente et al46 CLL 4‡ Tumor, blood Genome 1,038 (mean) 23 (mean) NOTCH1 and MYD88
mutations associated with
distinct clinical subgroups of
CLL

Abbreviations: AML, acute myeloid leukemia; CLL, chronic lymphocytic leukemia; GBM, glioblastoma multiforme; HCC, hepatocellular carcinoma; NA, not
applicable; PI3K, phosphoinositide 3-kinase.
ⴱ
Total of 20,661 protein coding genes were sequenced.
†Study also performed whole exome sequencing for 16 patients not reported in Table 3.
‡Study also performed targeted sequencing for 363 patients not reported in Table 3.

www.jco.org © 2012 by American Society of Clinical Oncology 5

Downloaded from jco.ascopubs.org on February 20, 2013. For personal use only. No other uses without permission.
Copyright © 2012 American Society of Clinical Oncology. All rights reserved.
 Tran et al

encompass other therapeutic strategies including regulation of the tions within coding regions, from which discovery of cancer genes and
immune system. In cancer immunotherapy, efforts aimed at person- potentially druggable genes will make real differences in can-
alizing treatment are investigating the potential insights gained from cer care.21,63,64
transcriptional profiling, including the exploration of the prognostic Drivers versus passengers. Within individual cancer genomes
and predictive impact of immune signatures in the tumor microenvi- sequenced thus far, hundreds to thousands of mutations are
ronment and of potential markers that may be associated with re- present.34-41 Assigning relevance to each is difficult. Many of these
sponse to immunotherapy. These approaches may also provide mutations are probably passenger mutations; that is, they were present
potential candidate tumor-associated antigens for the development of but irrelevant in the dominant cancer clone when critical mutations
cancer vaccines.55-59 leading to tumorigenesis were acquired, or they represent de novo
Our understanding of cancer is reliant not only on sequencing mutations that arose subsequently in rapidly dividing cancer cells. The
the cancer genome but also on identifying epigenetic and gene expres- greatest challenge in cancer genomic discovery is distinguishing be-
sion changes. Additionally, although the scope of this review is limited nign passenger mutations from those relevant to pathogenesis (ie,
primarily to genomics, the potential role of cancer proteomics in driver mutations). Driver mutations confer growth advantage and, by
PCM, as described by Hanash et al,60 must also be recognized. definition, reside in the subset of genes known as cancer genes.42,70
Knowing whether a mutation is recurrent and being aware of its
Conceptual Debates frequency in the examined sample can assist in differentiating between
Debate regarding cancer genomics occurs on multiple levels. drivers and passengers. However, the gold standard is functional val-
Weinberg61 raises concerns regarding the enormous amount of re- idation of the mutation in vitro and/or in vivo.
sources and researcher energy involved in sequencing entire cancer Passenger mutations are generally distributed randomly across
genomes with only modest dividends to date; Heng62 believes the the genome, whereas driver mutations cluster within cancer genes.42
current concept of cancer needs to be re-examined to ensure findings The likelihood of having identical mutations at the same position in
from ICGC projects are not wasted. Stratton et al42 note this debate is multiple samples is extremely low; therefore, the vast majority of
not new to genomics and liken it to arguments that occurred before mutations recurring in the same genomic regions are likely to be driver
the human genome was sequenced. They agree that the cost of se- mutations.42,69 Because many cancer genes seem to contribute to
quencing large numbers of cancer genomes may be astronomical and cancer development in only a small fraction of tumors (for example,
that findings are not entirely predictable; however, given that the only approximately 10% of colorectal cancers [CRCs] have a BRAF
human genome is finite, this is a deliverable project with a scope that mutation71), large sample sets must be analyzed to distinguish infre-
goes beyond identifying cancer genes. For example, as illustrated by quently mutated cancer genes from genes with random clusters of
the GBM and ovarian cancer genome projects, comprehensive cata- passenger mutations.42 To be confident of identifying a cancer gene
loguing of somatic mutations can generate insights into genetic pat- that is mutated in 5% of a particular type of cancer, hundreds of
terns that underlie phenotype, prognosis, and drug response.42 samples need to be sequenced.42
Mutation frequency refers to the percentage of DNA in which a
Challenges and Issues mutation is identified and can be used in statistical calculations that
Aside from the ongoing conceptual debate, we have identified differentiate drivers from passengers.42 In theory, heterozygous mu-
four major considerations in completing the molecular map of cancer: tations present in all cancer cells should have a frequency of 50%,
sequencing exomes versus whole genomes, differentiating between whereas homozygous germline variants should have a frequency of
driver and passenger mutations, integrating data across platforms, and 100%.72 Because tumors are typically contaminated by normal tissue
finding solutions to problems associated with multiple testing. (tumor stroma), mutational frequencies generated by NGS require a
Exome versus whole-genome sequencing. Exomes are protein- correction factor to determine the actual frequency of the mutation.35
coding genomic regions and constitute approximately 1% of the Although tumor heterogeneity can result in a wide spectrum of mu-
whole genome.63 Because of financial constraints, many cancer ge- tation frequencies, in general, driver mutations have frequencies at or
nome projects sequence exomes rather than whole genomes,21,38,63-65 lower than 50%.
yet whether this is the best approach remains debatable.21,64,66 Muta- Functional validation in tissue cultures or model organisms is the
tions within exomes are easier to interpret.66,67 For example, muta- gold standard for assessing mutation significance.66,69,70,72 This in-
tions that lead to a change in protein sequence can be easily volves assembling multiple mutations into a single cell or model or-
distinguished from noncoding variants. By contrast, there is less un- ganism and requires significant resources and time. Scott et al73
derstanding of the function of noncoding regions.42,68 On the other recently published an example of a complete genomics-to-function
hand, whole genome sequencing of a lung cancer cell line identified paradigm. They identified a genomic region at 5p13 that was com-
22,910 point mutations, but only 134 (0.6%) were in exomes.36 What monly amplified in several cancer types, including lung, ovarian, pros-
do the other 22,776 mutations represent? The answer to this question tate, and breast cancers and melanoma. An integrated analysis of this
is not known, but it can be argued that it will remain unanswered region pinpointed the Golgi-associated protein GOLPH3 for further
unless each one is systematically investigated.42,67 Mardis et al35 dis- study. In vitro and in vivo experiments involving knockdown and
covered a recurring nongenic mutation in sequencing an acute my- amplification of the GOLPH3 gene led to a linkage between GOLPH3
eloid leukemia genome. They contend that finding mutations in and the mammalian target of rapamycin, thus establishing GOLPH3
noncoding regions may greatly improve our understanding of cancer as a cancer gene.73 This supports the need for functional studies of
and uncover new mechanisms of cancer pathogenesis.69 The opposing genomic alterations discovered by sequencing.
view is that expending cost and time investigating mutations in non- Integrating data. NGS and the resulting rapid increase in
coding regions impedes progress in understanding the result of muta- genome-scale data production have created great challenges in data

6 © 2012 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Downloaded from jco.ascopubs.org on February 20, 2013. For personal use only. No other uses without permission.
Copyright © 2012 American Society of Clinical Oncology. All rights reserved.
 Cancer Genomics

integration. Data integration among multiple samples and techniques to point out that the identification of actionable and druggable genetic
is essential for making reliable inferences from genomics data,31,74,75 aberrations is only one of many means to increase the cancer thera-
because often, an individual technique tells only part of the story. For peutic armamentarium, and not all actionable and druggable cancer
example, genomic sequencing can identify structural variations, but targets are the products of genetic aberrations. Targets of cancer im-
only by adding a technique that assesses RNA levels, such as RNA-seq, munotherapy, such as CTLA4 in metastatic melanoma (inhibited by
can it be determined whether the structural variations affect the tran- ipilumumab to improve immune response), CD20 in diffuse large
scription levels of a gene. The major hurdles to integration are heter- B-cell lymphoma (targeted by rituximab), or prostatic acid phospha-
ogeneity of the experimental and analytic protocols, varying levels of tase in prostate cancer (targeted by immune cells induced by
data quality, and differences in data representation.31,74,75 In particu- sipuleucel-T), are not directly the results of genetic aberrations in these
lar, both the frequency and distribution of false positives and negatives cancers, yet they are druggable by biologic therapeutics and actionable
have to be understood for each data set before attempting to integrate because such targeted therapeutics can improve cancer outcomes.84-86
and interpret them. Data integration will remain challenging as more As the number of potentially actionable and druggable aberra-
cancer genomes are sequenced, and innovative bioinformatics strate- tions increases, the repertoire of novel therapeutics studied in early-
gies are needed to facilitate this process. phase clinical trials also increases. To validate the predictive value of
Multiple testing. The wealth of data produced by genomic stud- genetic aberrations, clinical trials need to be appropriately de-
ies using high-throughput technologies presents a unique set of chal- signed.87,88 Dancey et al89 have published guidelines for developing
lenges related to identification of true positives in the context of a large biomarker studies in early-phase clinical trials. They and others
number of statistical hypotheses and comparisons, so-called multiple suggest using molecular targets to guide patient selection for inves-
testing. This is particularly relevant in cancer genomics, where it is tigational targeted agents,90 although this could also apply to inves-
common for hundreds to thousands of genes to be scrutinized in tigational immunotherapeutic agents. Molecular profiling (MP) will
studies aimed at identifying prognostic, predictive, or diagnostic mo- enable enrollment onto these studies without the need for individual
lecular biomarkers in selected cancers. Naive application of standard tests for each genetic aberration, thus enhancing the efficiency of
hypothesis tests without adjustment for multiple testing in these situ- investigational drug development. For this reason, there is a burgeon-
ations will result in large numbers of nonreproducible false positives. ing need for MP in the clinic.
On the other hand, using multiple testing statistical methods to con-
trol false positives in situations where families of tests are performed MP
can greatly reduce the power to detect true discoveries.76 As the Three initial studies examined the feasibility of MP using high-
amount of genetic data available increases, so too will problems asso- throughput genotyping (Table 4). Thomas et al,22 Dias-Santagata et
ciated with multiple testing.77,78 al,23 and MacConaill et al24 collectively examined between 250 and
1,000 individual tumor specimens for 120 to 400 mutations in 13 to 33
known oncogenes and tumor suppressor genes. These studies found
TRANSLATION at least one mutation in 30% to 37% of tumor samples and concluded
that high-throughput genotyping enables sensitive and accurate on-
There is universal expectation that the technologic advances and the cogenic mutational profiling in human cancer specimens.
understanding of the molecular basis of cancer gained from studying More recently, studies have examined the feasibility of real-time
cancer genomes will be translated into benefits for patients with can- MP of tumors from actual patients and matching of the identified
cer. One of the most valuable translational approaches is to develop molecular profile with targeted treatments. Von Hoff et al91 con-
therapeutic agents that target genetic druggable aberrations discov- ducted a study of matching treatment to molecular profile in 86
ered by studying cancer genomes. Targeted agents that act selectively patients across nine different centers in the United States. Only 66
on cancer cells harboring these aberrations may provide a greater patients proceeded to MP, wherein 64 targets were examined using a
therapeutic effect than traditional cytotoxic agents or unmatched tar- combination of immunohistochemistry (IHC), FISH, and gene ex-
geted agents. pression microarrays. Each aberration was matched to a predefined
The druggable genome was first described by Hopkins et al79 in treatment. In 18 of 66 patients, they demonstrated progression-free
Nature in 2002. They defined a protein as druggable if a small molecule survival for matched treatment to be 1.3 times greater than that for the
could bind it at the required binding affinity. They noted a protein treatment patients had received immediately beforehand. Tsimberi-
may be druggable, but for it to be a potential drug target, it must be dou et al92 performed molecular analysis of 1,283 patients, with suc-
linked to disease.79 Although these definitions were not made with cess in 1,144 (89%). They used PCR, FISH, and IHC in examining for
cancer in mind, they can be extrapolated. By definition, actionable 11 separate aberrations. In their cohort, 40% of patients had at least
aberrations are disease modifying, but not all are druggable. one aberration. They matched each aberration to a targeted treatment
Currently, several genetic aberrations have been validated as ac- when available and demonstrated that patients who received matched
tionable and druggable. ERBB2 amplification in breast cancer, EGFR targeted therapy had better response rates and improved time to
mutations in non–small-cell lung cancer (NSCLC), and BRAF muta- treatment failure.
tions in melanoma are actionable and druggable, because they are While Von Hoff et al91 and Tsimberidou et al92 enrolled patients
prognostic, have their gene products targeted by molecularly targeted with any tumor type, cancer-specific genotyping studies have also
therapies, and are predictive of their benefit.26,80-83 Alternatively, al- been performed. Recently, Kim et al93 published a report on the
though IDH1 mutations in GBM are actionable as a prognostic bio- BATTLE (Biomarker-Integrated Approaches of Targeted Therapy for
marker, treatments targeting their gene products do not currently Lung Cancer Elimination) study, in which 255 patients with pre-
exist, and hence IDH1 mutations are not yet druggable.39 It is relevant treated metastatic NSCLC had their disease prospectively biopsied for

www.jco.org © 2012 by American Society of Clinical Oncology 7

Downloaded from jco.ascopubs.org on February 20, 2013. For personal use only. No other uses without permission.
Copyright © 2012 American Society of Clinical Oncology. All rights reserved.
 Tran et al

Table 4. Molecular Profiling in Clinical Tumor Samples

Samples
With at
Least One
Total Samples Genes Analyzed Mutation

Author No. Tumor Types No. Mutations/Aberrations Mutations Detected No. % Method
Dias-Santagata et al23 250 26 13 120 100 86 34 Multiplex PCR-based targeted SNP
analysis (SnaPshot; Applied
Biosystems, Carlsbad, CA)
MacConaill et al24 903 12 33 396 417 335 37 MassARRAY (Sequenom, San Diego,
CA) somatic mutation analysis
Thomas et al22 1,000 17 17 238 NA 298 30 MassARRAY (Sequenom) somatic
mutation analysis
Von Hoff et al91 86ⴱ 25 51† NA 84 98‡ IHC, FISH, and gene expression
microarray
Tsimberidou et al92 1,283 NA 11 12 NA NA 40‡ PCR, FISH, IHC

Abbreviations: IHC, immunohistochemistry; FISH, fluorescence in situ hybridization; NA, not applicable; PCR, polymerase chain reaction; SNP, single
nucleotide polymorphism.
ⴱ
Eighty-six evaluable patients.
†Thirteen also had protein expression examined by IHC.
‡Includes significant change in gene or protein expression.

evaluation of five biomarkers. They also matched each biomarker to a methylation, or other, should also be specified. Developmental thera-
predefined treatment and demonstrated a 46% 8-week disease control peutic agents are evolving constantly, such that any list of candidate
rate with this strategy. genes and corresponding aberrations can rapidly become out-
dated; thus, any gene list used in MP should be regularly reviewed
Princess Margaret Hospital–Ontario Institute for and revised.
Cancer Research Clinical Genomics Initiatives Gene selection for MP has focused on oncogenes, tumor sup-
The Princess Margaret Hospital (PMH) –University Health Net- pressor genes, and genomic stability genes. However, we believe that
work and the Ontario Institute for Cancer Research (OICR) have to fully implement PCM, germline genetic aberrations involved in
developed a genomic pathway strategy to systematically evaluate PCM drug metabolism should also be examined, because there is increasing
using NGS and high-throughput genotyping platforms. A feasibility evidence of their role in predicting the efficacy and toxicity of can-
study is currently enrolling patients. This study will perform MP using cer therapies.94-96
somatic mutation genotyping and targeted exome sequencing of can- The PMH-OICR–led feasibility study surveyed 19 experts in
cer genes through the MassARRAY (Sequenom) and PacBio RS (Pa- cancer genomics to create a candidate gene list for MP. Experts in-
cific Biosciences) platforms. Figure 1 outlines the study design. The cluded drug developers and genome scientists. An in-depth literature
addition of targeted sequencing allows detection of novel and poten- search of the PubMed, COSMIC, and Genecard databases led to the
tially actionable mutations in target genes, compared with genotyping selection of 194 genes with aberrations known to be important in
methods. Patients are being enrolled from five cancer centers tumorigenesis or drug metabolism. The survey asked each expert to
throughout Ontario, Canada, and will be observed for 2 years, with the assign an importance score to each gene: one, highest; two, interme-
impact of MP on treatment decisions reassessed at regular intervals. diate; three, lowest; and four, unknown. The mean score assigned to
Indicators of feasibility include a 21-day turnaround from consent to each gene was calculated, and genes were ranked from most to least
reporting of results and discovering actionable mutations in at least important. This list was used to generate a gene panel for MP. Inter-
30% of patients. During the course of this study, several issues and estingly, the survey identified significant differences in the way drug
challenges have been identified in translating cancer genomics into the developers and genome scientists assigned importance to genes (Ap-
clinic, some of which are discussed here. pendix Table A1, online only). Drug developers were significantly
Candidate gene selection. Candidate genes selected for MP are more likely to assign higher importance to genes targeted by estab-
influenced by the aims of MP, technology at hand, and treatment lished or investigational agents.97 Although cancer genomics has been
options available. The majority of MP studies presented in Table 4 the domain of genome scientists, this survey demonstrates that in-
performed extensive literature reviews in creating their gene lists, volvement of drug developers is crucial to ensure that a clinically
whereas Tsimberidou et al92 only selected genetic aberrations with relevant gene panel is created for MP.
validated molecular diagnostic tests in a Clinical Laboratory Improve- Profiling archival versus current tumor samples. Most patients
ment Amendments (CLIA) – certified laboratory. with cancer have archived formalin-fixed paraffin-embedded (FFPE)
Because MP is likely to be used for patient selection in early-phase tissue available for MP. However, it is unclear whether archived tumor
clinical trials in the near future, it is important that genes selected for tissue, generally from the primary tumor, accurately represents the
analysis are potential, albeit unvalidated, predictors of efficacy for current disease state, particularly if that is metastatic cancer that has
existing investigational therapies. The genetic aberration associated progressed through multiple treatments. Molecular discordance be-
with each selected gene, whether that is mutation, amplification, tween primary and metastatic disease does occur, but it seems to differ

8 © 2012 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Downloaded from jco.ascopubs.org on February 20, 2013. For personal use only. No other uses without permission.
Copyright © 2012 American Society of Clinical Oncology. All rights reserved.
 Cancer Genomics

Recording of
impact of
molecular profile
on treatment

Follow-up

Molecular profile Expert Validated Sanger

Clinician report generated mutations
panel sequencing

Treatment
Mutations
decisions
identified
Somatic mutation
Blood sample genotyping
Consent and DNA
Patient Tumor biopsy
screening extracted Targeted exome
Archived tumor
sequencing
Follow-up

Recording of
efficacy and toxicity
of matched
treatments

Fig 1. Princess Margaret Hospital–Ontario Institute for Cancer Research feasibility study of somatic mutation genotyping and targeted exome sequencing.
Patients sign informed consent to enter the study and undergo screening to ensure there are no contraindications to biopsy. A blood sample is collected for
germline DNA, a tumor biopsy is performed, and archived tumor is retrieved. After DNA extraction, somatic mutation genotyping is performed using the
MassARRAY platform (Sequenom, San Diego, CA), and targeted exome sequencing is performed using the PacBio RS platform (Pacific Biosciences, Menlo Park,
CA). If required, identified mutations are confirmed by Sanger or other validated methods in a College of American Pathologists/Clinical Laboratory Improvement
Amendments– certified laboratory. Validated mutations are reviewed by an expert panel, and a molecular profiling report is generated for the clinician. The study
aims to produce a report within 21 days of patient consent. The initial report outlines mutations in the fresh tumor biopsy. Delayed results from analysis of
archived tumor are provided in an amended report. The clinician reviews the report and makes treatment recommendations for the patient. Clinicians are asked
to record the impact of the molecular profiling report on their treatment recommendations. Patients are observed every 3 months for 2 years, and efficacy and
toxicity related to matched treatments are recorded.

between cancer types. In CRC, primary/metastatic concordance is clude optimization of tissue fixation and embedding methods. Both
high for KRAS and BRAF mutations.98,99 By contrast, in breast cancer, assist in obtaining clear histologic detail and facilitate elucidation of
primary/metastatic genotypic differences were observed in a WGS gene and protein expression profiles.20,105 Snap freezing and FFPE are
study, and reported discordance for both ERBB2 amplification and typical processing methods.105 Although FFPE tissue allows for histo-
estrogen receptor expression has not been insignificant.34,100 It re- logic examination, it does not prioritize preservation of DNA, RNA,
mains unclear whether the future of MP will require archived or and proteins.106,107 Traditionally, optimal molecular preservation in
current tumor samples; however, if current tumor samples are re- tissue has been achieved through snap freezing.105 However, many
quired, a fresh tumor biopsy is needed.101 Because biopsies are not centers are now comfortable with extracting DNA from FFPE sam-
without risk, noninvasive strategies are being explored, including cir- ples, and studies have shown that this can be done successfully.108
culating tumor cells and circulating DNA.102 The PMH-OICR feasi- These advances facilitate MP of archival specimens.
bility study is using specimens from fresh biopsies to perform MP. One technical challenge related to profiling of clinical samples is
However, archival specimens are also being collected and profiled so separating tumor from tumor stroma. A good-quality biopsy will
that primary/metastatic genotypes can be compared. generally ensure that the amount of necrotic tissue is minimal; how-
Although promising, novel methods of tumor sampling do not ever, cellular components from tumor stroma are sure to be included
address the problem of tumor heterogeneity. Tumor heterogeneity is if DNA is extracted from bulk tissue. In this situation, extracted DNA
defined as the simultaneous presence of multiple clonal subpopula- will be representative of both tumor cells and unwanted normal cells.
tions of tumor cells within a single neoplasm, and it represents a Laser-capture microdissection is a robust technique that can isolate
significant barrier to achieving PCM and a possible explanation for tumor cells,108-111 but the cost and time involved are not amenable to
mixed responses to targeted therapies.103,104 Performing multiple bi- real-time MP. Although less effective than laser-capture microdissec-
opsies at different sites is one potential solution to account for tumor tion for isolating tumor, macrodissection is more commonly used for
heterogeneity, but this is often impractical. Currently, there is no MP, because it is more economic and less time consuming. Alterna-
solution to this problem. tively, DNA can be extracted from bulk tumor, with the percentage of
Process optimization. The usefulness of tissue for MP is influ- tumor cells in the sample considered in evaluating the resultant mo-
enced significantly by specimen handling. Crucial components in- lecular profile. Regardless of how the DNA is acquired, there must be

www.jco.org © 2012 by American Society of Clinical Oncology 9

Downloaded from jco.ascopubs.org on February 20, 2013. For personal use only. No other uses without permission.
Copyright © 2012 American Society of Clinical Oncology. All rights reserved.
 Tran et al

a sufficient quantity for MP, and this is dependent on the amount of ing elements for both IHC and FISH: identification information,
viable tumor available within the sample. Hence, it is important that pathology processes, results, interpretation, and standardized
any sample used for MP be reviewed to assess sample quality, provide comment.119 Furthermore, it is recommended that ERBB2 testing
histopathologic confirmation of tumor, and determine the percentage be performed in a CLIA-certified, College of American Pathologists–
of tumor cells present. accredited laboratory.119
Technology validation. Many technologies are available for It remains unclear how MP results will be reported to clini-
high-throughput MP, including microarrays, mass spectrometry, cians, because methods for reporting are still in development. At
and NGS platforms. These platforms are able to analyze hundreds PMH-OICR, MP reports are generated after an expert panel meeting
to thousands of targets simultaneously. With every technology in which decisions are made regarding which mutations/aberra-
comes the potential for errors and biases, particularly resulting tions should be reported. The early experience of the PMH-OICR
from the long chain of decisions required in sampling, preprocess- expert panel (composed of clinical oncologists, pathologists, clin-
ing, processing, calibration, and analysis.112 Guidelines have been ical geneticists, ethicists, bioinformaticians, and genome scientists)
published regarding the minimum information needed for an is that discussion emanating from multiple perspectives is critical,
expression microarray experiment, and these have ensured process particularly when new germline or somatic mutations are discov-
standardization for this technology.113 For other platforms, there ered. Attached to each report is a document based on a literature
are no published guidelines at present. Ioannidis112 has opined that review that details the clinical significance, if any, of the muta-
“any molecular profile emerges eventually out of an abyss of exper- tion(s) discovered. Documents such as this or, alternatively, online
iments and analyses”112(p303) and suggests that standardization and resources such as MyCancerGenome (a freely available online
creation of a definitive and fixed profile is required before MP is PCM resource developed by the Vanderbilt–Ingram Cancer Center
ready for clinical decision making. The US Food and Drug Admin- that provides information regarding several important genetic
istration has also addressed concerns regarding reliability, preci- aberrations in select cancer types) are essential to ensure the clini-
sion, accuracy, and interlaboratory reproducibility of data derived cian is fully informed regarding the importance of any discov-
from MP.114 The CLIA program sets standards and issues certifi- ered mutation.120
cates for clinical laboratory testing on human specimens that pro- The Data Supplement provides an example of the PMH-OICR
vide information for the diagnosis, prevention, and treatment of MP report. Several considerations were taken into account in devel-
disease. Any laboratory that plans to perform MP to affect clinical oping this report, and although some relate to legal aspects, most are
decisions should be CLIA certified, ensuring accuracy, reliability, directed at assisting the clinician in interpreting the results (Table 5).
and timeliness of the test performed. Furthermore, although cur- As MP becomes more widespread, guidelines for reporting will be
rently only medical devices being evaluated in clinical trials are required. Furthermore, in keeping with guidelines for ERBB2 testing,
mandated to have an investigational device exemption by the US such reports should be generated by a CLIA/College of American
Food and Drug Administration, in the future, platforms used to Pathologists–approved molecular diagnostics laboratory.
identify integral biomarkers that affect clinical decisions on clinical
trials (eg, patient eligibility) are also likely to require this exemp-
tion as well.115
Once a platform is selected, it requires validation based on five Table 5. Issues to Be Addressed in Generating a Molecular Profiling Report
performance metrics: sequencing accuracy, variant accuracy, false- Category Questions
positive rate, false-negative rate, and variant discrepancy rate.116 Legal aspects Disclaimer
Harismendy et al116 evaluated three NGS platforms against Sanger Specification that report is for research use only
sequencing by sequencing a human genome using all four platforms. Process specific Listing of all mutations for which analysis occurred
Although they demonstrated that NGS identified more than 95% of Listing of analysis failures
variant alleles correctly, they also observed problems with systemic Description of platforms and methods used
Details of platform sensitivity and specificity
biases and data variability. Improvements to these technologies have
Laboratory CAP/CLIA certification
mitigated these issues, but centers using NGS for MP and clinical Mutation specific Inclusion of all mutations identified or only significant
decision making need to ensure that processes are optimized to min- mutations
imize potential problems. Even seemingly simple tests such as FISH Processes involved in deeming mutations significant
(eg, expert panel, functional validation)
and IHC require validation and thorough quality control before their
Mutation nomenclature (chromosome position v
use in biomarker-based studies, as learned from the lesson in ERCC1 amino acid change)
testing, where the antibody used for IHC staining was found to be not Inclusion of mutation frequency
specific for this marker after initial publications.117,118 At PMH-OICR, Clinical significance Inclusion of literature-based information regarding
of mutation clinical significance and known frequency of
somatic mutation genotyping and Sanger sequencing are performed mutation discovered
in a CLIA-certified laboratory, and the PacBio RS (Pacific Biosciences) Inclusion of preclinical data in addition to clinical data;
and MassARRAY (Sequenom) platforms were tested thoroughly and evaluation of level of evidence of clinical data
cross validated before the feasibility study opening. Listing of clinical trials involving novel agents
targeting identified mutation
Molecular profiling reports. Currently, clinicians request testing Listing of only clinical trials conducted in
for the few validated druggable and actionable genetic aberrations on geographically nearby centers
an as-needed basis, and results from molecular diagnostic laboratories Abbreviations: CAP, College of American Pathologists; CLIA, Clinical Laboratory
are reported back to clinicians in a standardized format. For ERBB2 Improvement Amendments.
testing in breast cancer, published guidelines outline required report-

10 © 2012 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Downloaded from jco.ascopubs.org on February 20, 2013. For personal use only. No other uses without permission.
Copyright © 2012 American Society of Clinical Oncology. All rights reserved.
 Cancer Genomics

Ethics. Privacy, confidentiality, and potential for subsequent industry, which must accept that its potential market will be reduced
discrimination (from increased cancer risk identified by GWAS) have as treatment becomes more personalized.112,129
been identified as major considerations in cancer genomics. Addition-
ally, three major ethical considerations associated with WGS have
been identified by McGuire et al121: identifying circumstances when DISCUSSION
research results are disclosed to participants, identifying obligations
that are owed to participants’ close genetic relatives, and deciding how The potential impact of cancer genomics is enormous. Ongoing
future uses of samples and data taken from WGS are handled. Al- technologic advances and falling costs of genome sequencing are
though McGuire et al address ethical issues associated with sequencing increasing the rate at which cancer sequencing projects can be
germline DNA, these issues also have implications for MP of tumor completed. As a result, identification, validation, and functional
DNA, because germline DNA is usually analyzed simultaneously. investigation of genetic aberrations in cancer are being pursued,
McGuire et al121 recognize that the volume, complexity, and and the list of genes of interest is expanding continually. Driven by
clinical uncertainty of data generated from WGS can make commu- the three pillars of cancer genomics—technology, discovery, and
nication of research results challenging and recommend that expertise translation—PCM is now within reach. However, before we are
is needed for interpretation and to ensure adequate understanding of able to translate the knowledge gained into clinical benefit that
health and social implications. Two other studies have examined the enhances patient care, key hurdles must be overcome. Although
ethics behind reporting genetic research results to study participants bioinformatic analysis and stringent quality control have reduced
and concluded that only results that are associated with sufficient risk errors associated with sequencing and genotyping, the reliability
and that have established clinical utility should be communicated back and accuracy of novel technologies remain potential problems.
to patients.122,123 Furthermore, McGuire et al suggest that only vali- From a discovery viewpoint, finding solutions to problems associ-
dated data approved by credentialed clinical laboratories be included ated with multiple testing and data integration is crucial to make
in health records. This approach has been adopted in the PMH-OICR reliable inferences from genomics data. Finally, although matching
feasibility study. treatments to novel biomarkers is the crux of PCM, key challenges
Clinical impact. The presence of ERBB2 amplification in associated with tissue processing and tumor heterogeneity must be
breast cancer, EGFR mutation in NSCLC, and KRAS mutation in recognized and addressed. Ultimately, prospective clinical valida-
CRC, to name a few, are genetic aberrations known to influence tion is needed to confirm that PCM provides cost-effective benefit
clinical decisions in medical oncology.26,80-82,124,125 In the near against our current standard approaches.
term, it is unlikely that a more comprehensive molecular profile
will influence the prescription of regulatory agency–approved
treatments, given the currently small number of validated predic- AUTHORS’ DISCLOSURES OF POTENTIAL CONFLICTS
OF INTEREST
tive biomarkers. However, the evolving nature of novel therapies in
early-phase clinical trials makes it attractive to use MP to select
Although all authors completed the disclosure declaration, the following
patients for clinical trials, particularly because matched treatments author(s) indicated a financial or other interest that is relevant to the subject
seem to provide clinical benefit.91,92 matter under consideration in this article. Certain relationships marked
Past studies of MP have demonstrated that approximately 30% with a “U” are those for which no compensation was received; those
of tumors profiled have actionable mutations (Table 4).22,23,24 relationships marked with a “C” were compensated. For a detailed
Whether knowledge of these aberrations will have a clinical impact is description of the disclosure categories, or for more information about
unknown. Although many novel agents are in clinical trials, access to ASCO’s conflict of interest policy, please refer to the Author Disclosure
Declaration and the Disclosures of Potential Conflicts of Interest section in
these early-phase studies can be difficult for those who practice outside Information for Contributors.
of large-volume cancer centers. Thus, for a true clinical impact, MP Employment or Leadership Position: None Consultant or Advisory
must lead to identification of actionable genetic aberrations in a sub- Role: Benjamin G. Neel, Kolltan Pharmaceuticals (C), Novartis (C)
stantial proportion of patients, and novel agents targeting these aber- Stock Ownership: Benjamin G. Neel, Kolltan Pharmaceuticals
rations must exist and be accessible through clinical trials at a Honoraria: None Research Funding: None Expert Testimony: None
reasonable distance to patients. Furthermore, results of MP should be Other Remuneration: None
communicated to clinicians within a reasonable timeframe, so pa-
tients remain well enough for clinicians to act on actionable aberra- AUTHOR CONTRIBUTIONS
tions. This is particularly important because patients with advanced
cancers can deteriorate rapidly. The PMH-OICR study timeframe of
Conception and design: Ben Tran, Janet E. Dancey, Nicole Onetto,
21 days from consent to results is consistent with other genotyping Thomas J. Hudson, Benjamin G. Neel, Lillian L. Siu
studies, which have reported median turnaround times of 20 to Financial support: Thomas J. Hudson, Benjamin G. Neel, Lillian L. Siu
30 days.126,127 Administrative support: Janet E. Dancey, Lillian L. Siu
The potential impact of MP is not limited to clinical decision Provision of study materials or patients: All authors
making. MP also provides an opportunity to link pathway deregula- Collection and assembly of data: Ben Tran, Suzanne Kamel-Reid,
tion with potential therapeutic strategies through biomarker valida- Andrew M.K. Brown, Tong Zhang, Patricia Shaw, Lillian L. Siu
Data analysis and interpretation: Ben Tran, Suzanne Kamel-Reid, John
tion.128 For MP to evolve from an interesting scientific concept into a D. McPherson, Philippe L. Bedard, Andrew M.K. Brown, Tong Zhang,
clinical tool, several issues must be addressed and resolved: impact on Patricia Shaw, Lincoln Stein, Benjamin G. Neel, Lillian L. Siu
clinical decision making, impact on drug development and biomarker Manuscript writing: All authors
validation, cost effectiveness, and buy-in from the pharmaceutical Final approval of manuscript: All authors

www.jco.org © 2012 by American Society of Clinical Oncology 11

Downloaded from jco.ascopubs.org on February 20, 2013. For personal use only. No other uses without permission.
Copyright © 2012 American Society of Clinical Oncology. All rights reserved.
 Tran et al

22. Thomas RK, Baker AC, Debiasi RM, et al: 41. Ley TJ, Mardis ER, Ding L, et al: DNA
REFERENCES High-throughput oncogene mutation profiling in hu- sequencing of a cytogenetically normal acute my-
man cancer. Nature Genet 39:347-351, 2007 eloid leukaemia genome. Nature 456:66-72, 2008
1. National Cancer Institute: Dictionary of can- 23. Dias-Santagata D, Akhavanfard S, David SS, 42. Stratton MR, Campbell PJ, Futreal PA: The
cer terms: Personalized medicine. http://www.cancer et al: Rapid targeted mutational analysis of human cancer genome. Nature 458:719-724, 2009
.gov/dictionary?CdrID⫽561717 tumours: A clinical platform to guide personalized 43. International Cancer Genome Consortium:
2. National Cancer Institute: Office of Cancer cancer medicine. EMBO Mol Med 2:146-158, 2010 International network of cancer genome projects.
Genomics. http://ocg.cancer.gov/ 24. MacConaill LE, Campbell CD, Kehoe SM, et Nature 464:993-998, 2010
3. Sjöblom T, Jones S, Wood LD, et al: The al: Profiling critical cancer gene mutations in clinical 44. Hirschhorn JN, Daly MJ: Genome-wide as-
consensus coding sequences of human breast and tumor samples. PloS One 4:e7887, 2009 sociation studies for common diseases and complex
colorectal cancers. Science 314:268-274, 2006 25. Sauter G, Lee J, Bartlett JM, et al: Guidelines traits. Nature Rev Genet 6:95-108, 2005
4. Metzker ML: Sequencing technologies: The for human epidermal growth factor receptor 2 test- 45. Cancer Genome Atlas Research Network:
next generation. Nat Rev Genet 11:31-46, 2009 ing: Biologic and methodologic considerations. Integrated genomic analyses of ovarian carcinoma.
5. Shendure J, Ji H: Next-generation DNA se- J Clin Oncol 27:1323-1333, 2009 Nature 474:609-615, 2011
quencing. Nat Biotechnol 26:1135-1145, 2009 26. Slamon DJ, Leyland-Jones B, Shak S, et al: 46. Puente XS, Pinyol M, Quesada V, et al:
6. International Human Genome Consortium: Use of chemotherapy plus a monoclonal antibody Whole-genome sequencing identifies recurrent mu-
Finishing the euchromatic sequence of the human against HER2 for metastatic breast cancer that tations in chronic lymphocytic leukaemia. Nature
genome. Nature 431:931-945, 1992 overexpresses HER2. N Engl J Med 344:783-792, 475:101-105, 2011
7. Robison K: Application of second-generation 2001 47. Jones S, Zhang X, Parsons DW, et al: Core
sequencing to cancer genomics. Brief Bioinform 27. Bang YJ, Van Cutsem E, Feyereislova A, et signaling pathways in human pancreatic cancers
11:524-534, 2010 al: Trastuzumab in combination with chemotherapy revealed by global genomic analyses. Science 321:
8. Lin J, Li M: Molecular profiling in the age of versus chemotherapy alone for treatment of HER2- 1801-1806, 2008
cancer genomics. Exp Rev Mol Diagn 8:263-276, positive advanced gastric or gastro-oesophageal 48. Chapman MA, Lawrence MS, Keats JJ, et al:
2008 junction cancer (ToGA): A phase 3, open-label, ran- Initial genome sequencing and analysis of multiple
9. Ajay SS, Parker SC, Abaan HO, et al: Accu- domised controlled trial. Lancet 376:687-697, 2010 myeloma. Nature 471:467-472, 2011
rate and comprehensive sequencing of personal 28. Davies JJ, Wilson IM, Lam WL: Array CGH 49. Totoki Y, Tatsuno K, Yamamoto S, et al:
genomes. Genome Res 21:1498-1505, 2011 High-resolution characterization of a hepatocellular
technologies and their applications to cancer ge-
10. Zhao J, Grant SF: Advances in whole ge- carcinoma genome. Nat Genet 43:464-469, 2011
nomes. Chromosome Res 13:237-248, 2005
nome sequencing technology. Curr Pharm Biotech- 50. Esteller M: Epigenetics in cancer. N Engl
29. Andrews J, Kennette W, Pilon J, et al: Multi-
J Med 358:1148-1159, 2008
nol 12:293-305, 2011 platform whole-genome microarray analyses refine
51. Kim K, Zakharkin SO, Allison DB: Expecta-
11. Eid J, Fehr A, Gray J, et al: Real-time DNA the epigenetic signature of breast cancer metastasis
tions, validity, and reality in gene expression profil-
sequencing from single polymerase molecules. Sci- with gene expression and copy number. PloS One
ing. J Clin Epidemiol 63:950-959, 2010
ence 323:133-138, 2009 5:e8665, 2010
52. Perou CM, Sørlie T, Eisen MB, et al: Molec-
12. Rohde H, Qin J, Cui Y, et al: Open-source 30. Pop M, Salzberg SL: Bioinformatics chal-
ular portraits of human breast tumours. Nature 406:
genomic analysis of shiga-toxin-producing E. coli lenges of new sequencing technology. Trends
747-752, 2000
O104:H4. N Engl J Med 365:718-724, 2011 Genet 24:142-149, 2008
53. Vijver MJ van de, He YD, Veer LJ van’t, et al:
13. Twyman RM, Primrose SB: Techniques pat- 31. Teufel A, Krupp M, Weinmann A, et al:
A gene-expression signature as a predictor of sur-
ents for SNP genotyping. Pharmacogenomics 4:67- Current bioinformatics tools in genomic biomedical
vival in breast cancer. N Engl J Med 347:1999-2009,
79, 2003 research. Int J Mol Med 17:967-973, 2006
2002
14. Isler JA, Vesterqvist OE, Burczynski ME: 32. Saeys Y, Inza I, Larrañaga P: A review of
54. Paik S, Shak S, Tang G, et al: A multigene
Analytical validation of genotyping assays in the feature selection techniques in bioinformatics.
assay to predict recurrence of tamoxifen-treated,
biomarker laboratory. Pharmacogenomics 8:353- Bioinformatics 23:2507-2517, 2007
node-negative breast cancer. N Engl J Med 351:
368, 2007 33. Pandya GA, Holmes MH, Sunkara S, et al: A
2817-2826, 2004
15. Brion M, Quintela I, Sobrino B, et al: New bioinformatic filter for improved base-call accuracy
55. Wang E, Panelli MC, Marincola FM: Gene
technologies in the genetic approach to sudden and polymorphism detection using the Affymetrix profiling of immune responses against tumors. Cur
cardiac death in the young. Forensic Sci Int 203:15- GeneChip whole-genome resequencing platform. Opin Immunol 17:423-427, 2005
24, 2010 Nucleic Acids Res 35:e148, 2007 56. Pardoll DM: Spinning molecular immunology
16. Tsuchihashi Z, Dracopoli N: Progress in high 34. Ding L, Ellis MJ, Li S, et al: Genome remod- into successful immunotherapy. Nature Rev Immu-
throughput SNP genotyping methods. Biosystems elling in a basal-like breast cancer metastasis and nol 2:227-238, 2002
2:103-110, 2002 xenograft. Nature 464:999-1005, 2010 57. Galon J, Costes A, Sanchez-Cabo F, et al:
17. Pati N, Schowinsky V, Kokanovic O, et al: A 35. Mardis ER, Ding L, Dooling DJ, et al: Type, density, and location of immune cells within
comparison between SNaPshot, pyrosequencing, Recurring mutations found by sequencing an human colorectal tumors predict clinical outcome.
and biplex invader SNP genotyping methods: Accu- acute myeloid leukemia genome. N Engl J Med Science 313:1960-1964, 2006
racy, cost, and throughput. J Biochem Biophys 361:1058-1066, 2009 58. Bedognetti D, Wang E, Sertoli MR, et al:
Methods 60:1-12, 2004 36. Pleasance ED, Stephens PJ, O’Meara S, et Gene-expression profiling in vaccine therapy and
18. Koumi P, Green HE, Hartley S, et al: Evalua- al: A small-cell lung cancer genome with complex immunotherapy for cancer. Exp Rev Vaccines 9:555-
tion and validation of the ABI 3700, ABI 3100, and signatures of tobacco exposure. Nature 463:184- 565, 2010
the MegaBACE 1000 capillary array electrophoresis 190, 2010 59. Coppola D, Nebozhyn M, Khalil F, et al:
instruments for use with short tandem repeat mic- 37. Pleasance ED, Cheetham RK, Stephens PJ, Unique ectopic lymph node-like structures present
rosatellite typing in a forensic environment. Electro- et al: A comprehensive catalogue of somatic muta- in human primary colorectal carcinoma are identified
phoresis 25:2227-2241, 2004 tions from a human cancer genome. Nature 463: by immune gene array profiling. Am J Path 179:37-
19. Curtis C, Lynch AG, Dunning MJ, et al: The 191-196, 2010 45, 2011
pitfalls of platform comparison: DNA copy number 38. Cancer Genome Atlas Research Network: 60. Hanash S, Taguchi A: The grand challenge to
array technologies assessed. BMC Genomics 10: Comprehensive genomic characterization defines decipher the cancer proteome. Nature Rev Cancer
588, 2009 human glioblastoma genes and core pathways. Na- 10:652-660, 2010
20. Voidonikolas G, Kreml SS, Chen C, et al: ture 455:1061-1068, 2008 61. Weinberg R: Point: Hypotheses first. Nature
Basic principles and technologies for deciphering 39. Parsons DW, Jones S, Zhang X, et al: An 464:678, 2010
the genetic map of cancer. World J Surg 33:615- integrated genomic analysis of human glioblastoma 62. Heng HH: Cancer genome sequencing: The
629, 2009 multiforme. Science 321:1807-1812, 2008 challenges ahead. Bioessays 29:783-794, 2007
21. Teer JK, Mullikin JC: Exome sequencing: 40. Berger MF, Lawrence MS, Demichelis F, et 63. Ng SB, Turner EH, Robertson PD, et al:
The sweet spot before whole genomes. Hum Mol al: The genomic complexity of primary human pros- Targeted capture and massively parallel sequencing
Genet 19:145-151, 2010 tate cancer. Nature 470:214-220, 2011 of 12 human exomes. Nature 461:272-276, 2009

12 © 2012 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Downloaded from jco.ascopubs.org on February 20, 2013. For personal use only. No other uses without permission.
Copyright © 2012 American Society of Clinical Oncology. All rights reserved.
 Cancer Genomics

64. Maher B: Exome sequencing takes centre 87. Gonzalez-Angulo AM, Hennessy BT, Mills BRAF tumors. J Clin Oncol 28:15s, 2010 (suppl;
stage in cancer profiling. Nature 459:146-147, 2009 GB: Future of personalized medicine in oncology: A abstr 3534)
65. Cirulli ET, Singh A, Shianna KV, et al: Screen- systems biology approach. J Clin Oncol 28:2777- 105. Leiva IM, Emmert-Buck MR, Gillespie JW:
ing the human exome: A comparison of whole 2783, 2010 Handling of clinical tissue specimens for molecular
genome and whole transcriptome sequencing. Ge- 88. Macconaill LE, Garraway LA: Clinical implica- profiling studies. Curr Issues Mol Biol 5:27-35, 2003
nome Biol 11:R57, 2010 tions of the cancer genome. J Clin Oncol 28:5219- 106. Klimecki WT, Futscher BW, Dalton WS: Ef-
66. Diamandis EP, Hudson T, Kallioniemi O, et al: 5228, 2010 fects of ethanol and paraformaldehyde on RNA yield
Cancer genomes. Clin Chem 56:1660-1664, 2010 89. Dancey JE, Dobbin KK, Groshen S, et al: and quality. BioTechniques 16:1021-1023, 1994
67. Strausberg RL, Simpson AJ: Whole-genome Guidelines for the development and incorporation of 107. Ben-Ezra J, Johnson DA, Rossi J, et al: Effect
cancer analysis as an approach to deeper under- biomarker studies in early clinical trials of novel of fixation on the amplification of nucleic acids from
standing of tumour biology. Br J Cancer 102:243- agents. Clin Cancer Res 16:1745-1755, 2010 paraffin-embedded material by the polymerase
248, 2010 90. Doroshow JH: Selecting systemic cancer chain reaction. J Histochem Cytochem 39:351-354,
68. Ng PC, Levy S, Huang J, et al: Genetic therapy one patient at a time: Is there a role for 1991
variation in an individual human exome. PLoS Genet molecular profiling of individual patients with ad- 108. Coura R, Prolla JC, Meurer L, et al: An
4:e1000160, 2008 vanced solid tumors? J Clin Oncol 28:4869-4871, alternative protocol for DNA extraction from forma-
69. Walter MJ, Graubert TA, Dipersio JF, et al: 2010 lin fixed and paraffin wax embedded tissue. J Clin
Next-generation sequencing of cancer genomes: 91. Von Hoff DD, Stephenson JJ Jr, Rosen P, et Path 58:894-895, 2005
Back to the future. Per Med 6:653, 2009 al: Pilot study using molecular profiling of patients 109. Bonner RF: Laser capture microdissection:
70. Bell DW: Our changing view of the genomic tumors to find potential targets and select treat- Molecular analysis of tissue. Science 278:1481-
landscape of cancer. J Path 220:231-243, 2010 ments for their refractory cancers. J Clin Oncol 1483, 1997
71. Tran B, Kopetz S, Tie J, et al: Impact of BRAF 28:4877-4883, 2010 110. Hunt JL, Finkelstein SD: Microdissection
mutation and microsatellite instability on the pattern 92. Tsimberidou AM, Iskander NG, Hong DS, et techniques for molecular testing in surgical pathol-
of metastatic spread and prognosis in metastatic al: Personalized medicine in a phase I clinical trials ogy. Arch Path Lab Med 128:1372-1378, 2004
colorectal cancer. Cancer [epub ahead of print on program: The M. D. Anderson Cancer Center initia- 111. Cardoso J, Molenaar L, de Menezes RX, et
March 31, 2011] tive. J Clin Oncol 29:18s, 2011 (suppl; abstr al: Genomic profiling by DNA amplification of laser
72. Mardis ER, Wilson RK: Cancer genome se- CRC2500) capture microdissected tissues and array CGH. Nu-
quencing: A review. Hum Mol Genet 18:R163-R168, 93. Kim ES, Herbst RS, Wistuba II, et al: The cleic Acids Res 32:e146, 2004
2009 BATTLE Trial: Personalizing Therapy for Lung Can-
112. Ioannidis JP: Is molecular profiling ready for
73. Scott KL, Kabbarah O, Liang MC, et al: cer. Cancer Discov 1:44-53, 2011
use in clinical decision making? Oncologist 12:301-
GOLPH3 modulates mTOR signalling and rapamycin 94. Ross CJ, Katzov-Eckert H, Dubé MP, et al:
311, 2007
sensitivity in cancer. Nature 459:1085-1090, 2009 Genetic variants in TPMT and COMT are associated
113. Brazma A, Hingamp P, Quackenbush J, et al:
74. Stein LD: Integrating biological databases. with hearing loss in children receiving cisplatin chem-
Minimum information about a microarray experi-
Nat Rev Genet 4:337-345, 2003 otherapy. Nature Genet 41:1345-1349, 2009
ment (MIAME): Toward standards for microarray
75. Kemmeren P, Holstege FC: Integrating func- 95. de Haas EC, Zwart N, Meijer C, et al: Varia-
data. Nat Genet 29:365-371, 2001
tional genomics data. Biochem Soc Trans 31:1484- tion in bleomycin hydrolase gene is associated with
114. Ardekani AM, Petricoin EF, Hackett JL: Mo-
1487, 2003 reduced survival after chemotherapy for testicular
lecular diagnostics: An FDA perspective. Expert Rev
76. Manly KF, Nettleton D, Hwang JT: Genom- germ cell cancer. J Clin Oncol 26:1817-1823, 2008
Mol Diagn 3:129-140, 2003
ics, prior probability, and statistical tests of multiple 96. O’Dwyer PJ, Catalano RB: Uridine diphos-
115. US Food and Drug Administration: Device
hypotheses. Genome Res 14:997-1001, 2004 phate glucuronosyltransferase (UGT) 1A1 and irino-
advice: Investigational device exemption (IDE). http://
77. Nadon R, Shoemaker J: Statistical issues tecan: Practical pharmacogenomics arrives in cancer
www.fda.gov/MedicalDevices/DeviceRegulationand
with microarrays: Processing and analysis. Trends therapy. J Clin Oncol 24:4534-4538, 2006
Guidance/HowtoMarketYourDevice/Investigational
Genet 18:265-271, 2002 97. Tran B, Dancey JE, Kamel-Reid S, et al:
DeviceExemptionIDE/default.htm
78. Martin LJ, Woo JG, Avery CL, et al: Multiple Generation of a panel of druggable mutation cancer
testing in the genomics era: Findings from Genetic 116. Harismendy O, Ng PC, Strausberg RL, et al:
genes for molecular profiling in a feasibility study of
Analysis Workshop 15, Group 15. Genet Epidemiol somatic mutation genotyping and targeted exome Evaluation of next generation sequencing platforms
31:124-131, 2007 sequencing: OICR-PMH study. Presented at the for population targeted sequencing studies. Ge-
79. Hopkins AL, Groom CR: The druggable ge- European Organisation for Research and Treatment nome Biol 10:R32, 2009
nome. Nat Rev Drug Discov 1:727-730, 2002 of Cancer–National Cancer Institute–American As- 117. Bhagwat NR, Roginskaya VY, Acquafondata
80. Geyer CE, Forster J, Lindquist D, et al: sociation for Cancer Research Annual Meeting on MB, et al: Immunodetection of DNA repair endonu-
Lapatinib plus capecitabine for HER2-positive ad- Molecular Markers in Cancer, Brussels, Belgium, clease ERCC1-XPF in human tissue. Cancer Res
vanced breast cancer. N Engl J Med 355:2733-2743, October 27-29, 2011 69:6831-6838, 2009
2006 98. Vakiani E, Janakiraman M, Shen R, et al: 118. Olaussen KA, Fouret P, Kroemer G: ERCC1-
81. Mok TS, Wu YL, Thongprasert S, et al: Comparative genomic analysis of primary versus specific immunostaining in non-small-cell lung can-
Gefitinib or carboplatin-paclitaxel in pulmonary ade- metastasis in colorectal carcinomas. J Clin Oncol cer. N Engl J Med 357:1559-1561, 2007
nocarcinoma. N Engl J Med 361:947-957, 2011 29:15s, 2011 (suppl; abstr 10500) 119. Wolff AC, Hammond ME, Schwartz JN, et al:
82. Kim E, Hirsh V, Mok T, et al: Gefitinib versus 99. Tie J, Gibbs P, Lipton L, et al: Optimizing American Society of Clinical Oncology/College of
docetaxel in previously treated non-small-cell lung targeted therapeutic development: Analysis of American Pathologists guideline recommendations
cancer (INTEREST): A randomised phase III trial. a colorectal cancer patient population with the for human epidermal growth factor receptor 2 test-
Lancet 372:1809-1818, 2008 BRAFV600E mutation. Int J Cancer 128:2075- ing in breast cancer. J Clin Oncol 25:118-145, 2007
83. Chapman PB, Hauschild A, Robert C, et al: 2084, 2010 120. My Cancer Genome: Genetically informed
Improved survival with vemurafenib in melanoma 100. Amir E, Miller N, Geddie W, et al: Prospec- cancer medicine. http://www.mycancergenome.org
with BRAF V600E mutation. N Engl J Med 364: tive study evaluating the impact of tissue confirma- 121. McGuire AL, Caulfield T, Cho MK: Research
2507-2516, 2011 tion of metastatic disease in patients with breast ethics and the challenge of whole-genome sequenc-
84. Coiffier B, Lepage E, Briere J, et al: CHOP cancer. J Clin Oncol doi: 10.1200/JCO.2010.33.5232 ing. Nat Rev Genet 9:152-156, 2008
chemotherapy plus rituximab compared with CHOP 101. Janes CH, Lindor KD: Outcome of patients 122. Fabsitz RR, McGuire A, Sharp RR, et al:
alone in elderly patients with diffuse large-B-cell hospitalized for complications after outpatient liver Ethical and practical guidelines for reporting genetic
lymphoma. N Engl J Med 346:235-242, 2002 biopsy. Ann Intern Med 118:96-98, 1993 research results to study participants: Updated
85. Hodi FS, O’Day SJ, McDermott DF, et al: 102. Sawyers CL: The cancer biomarker problem. guidelines from a National Heart, Lung, and Blood
Improved survival with ipilimumab in patients with Nature 452:548-552, 2008 Institute Working Group. Circ Cardiovasc Genet
metastatic melanoma. N Engl J Med 363:711-723, 103. Li J, Wang K, Jensen TD, et al: Tumor 3:574-580, 2010
2010 heterogeneity in neoplasms of breast, colon, and 123. Bookman EB, Langehorne AA, Eckfeldt JH,
86. Kantoff PW, Higano CS, Shore ND, et al: skin. BMC Res Notes 3:321, 2010 et al: Reporting genetic results in research studies:
Sipuleucel-T immunotherapy for castration-resistant 104. Kopetz S, Desai J, Chan E, et al: PLX4032 in Summary and recommendations of an NHLBI work-
prostate cancer. N Engl J Med 363:411-422, 2010 metastatic colorectal cancer patients with mutant ing group. Am J Med Genet 140:1033-1040, 2006

www.jco.org © 2012 by American Society of Clinical Oncology 13

Downloaded from jco.ascopubs.org on February 20, 2013. For personal use only. No other uses without permission.
Copyright © 2012 American Society of Clinical Oncology. All rights reserved.
 Tran et al

124. Amado RG, Wolf M, Peeters M, et al: Wild- 126. Ortiz TM, Joshi VA, Heon S, et al: The (NSCLC) in clinical practice. J Clin Oncol 29:15s, 2011
type KRAS is required for panitumumab efficacy in introduction of systematic genomic testing for pa- (suppl; abstr 7518)
patients with metastatic colorectal cancer. J Clin tients with non-small cell lung cancer (NSCLC) at 128. Bild AH, Potti A, Nevins JR: Linking onco-
Oncol 26:1626-1634, 2008 Dana-Farber Cancer Institute (DFCI). J Clin Oncol genic pathways with therapeutic opportunities. Na-
125. Karapetis CS, Khambata-Ford S, Jonker DJ, et al: 29:15s, 2011 (suppl; abstr 7517) ture Rev Cancer 6:735-741, 2006
K-ras mutations and benefit from cetuximab in advanced 127. Sequist LV, Heist RS, Shaw AT, et al: 129. Harris TJ, McCormick F: The molecular pathol-
colorectal cancer. N Engl J Med 359:1757-1765, 2008 SNaPshot genotyping of non-small cell lung cancers ogy of cancer. Nat Rev Clin Oncol 7:251-265, 2010

■ ■ ■

Acknowledgment
We thank the numerous clinicians, scientists, technicians, and trainees who contributed to designing and establishing the Princess Margaret
Hospital (PMH) –Ontario Institute for Cancer Research (OICR) Genome Pathway Strategy described in this article and acknowledge support
by the PMH Foundation, Cancer Care Ontario, and OICR through the Ontario Ministry of Research.

Appendix

Glossary Terms

Actionable aberrations: Any aberration (germline or so- Germline mutation: An inherited variation in the lineage of germ
matic) that may impact cancer management through diagnostic, cells. Germline mutations can be passed on to offspring.
prognostic, and/or predictive implications.

Driver mutation: Driver mutations are those that are caus- Passenger mutation: A somatic mutation present in cancer with
ally implicated in oncogenesis or tumor survival. Such mutations no bearing on tumor biology.
have been positively selected during carcinogenesis and often
show a recurrent pattern within or across tumor types. This is in
Sequencing: A laboratory process that determines the nucleotide
contrast with passenger events, which arise from the background
sequence of DNA (can involve the whole genome or whole exome or be
mutation rate and do not contribute to oncogenesis.
targeted to as little as one coding sequence). Unlike somatic mutation
Druggable aberrations: Any actionable aberration that can genotyping, sequencing can detect previously unknown somatic
be targeted by novel therapeutics and thereby change the natural mutations.
history of the disease.

Exome: Part of the genome formed by genes that code for pro- SNP (single nucleotide polymorphism): Genetic polymor-
teins and other functional gene products (known as exons). phisms are natural variations in the genomic DNA sequence present in
greater than 1% of the population, with SNP representing DNA varia-
Genome: The complete set of genetic material. tions in a single nucleotide. SNPs are being widely used to better under-
stand disease processes, thereby paving the way for genetic-based
Genotyping: The process used for obtaining the genotype of a diagnostics and therapeutics.
given gene. From a somatic point of view (within a tumor), geno-
typing can be used to identify a predetermined genetic aberra-
tion, such as somatic mutations, copy number variations, gene Somatic mutation: A change in the genotype of a cancer cell. This
expression changes, and/or DNA methylation. Genotyping iden- is distinguished from a germline mutation, which is a change in the ge-
tifies only predetermined aberrations, with all other aberrations notype of all the normal cells in a patient’s body. Germline mutations
being effectively invisible. may be passed to offspring, but somatic mutations may not.

14 © 2012 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Downloaded from jco.ascopubs.org on February 20, 2013. For personal use only. No other uses without permission.
Copyright © 2012 American Society of Clinical Oncology. All rights reserved.
 Cancer Genomics

Table A1. Top 50 Genes With Aberrations Important in Tumor Biologyⴱ

Overall (n ⫽ 19) Drug Developers (n ⫽ 10) Genome Scientists (n ⫽ 9)

Rank Gene Mean SD Gene Mean SD Gene Mean SD

1 EGFR 1.05 0.23 EGFR 1.00 0.00 BRAF 1.00 1.11
2 BRAF 1.11 0.32 ERBB2 1.00 0.00 BRCA1 1.00 1.11
3 KIT 1.11 0.32 KIT 1.00 0.00 BRCA2 1.00 1.11
4 BRCA1 1.16 0.37 ALK 1.10 0.32 EGFR 1.00 1.11
5 BRCA2 1.16 0.37 BCR 1.10 0.32 KIT 1.00 1.22
6 ERBB2 1.16 0.37 BRAF 1.10 0.32 KRAS 1.00 1.22
7 KRAS 1.16 0.37 KRAS 1.10 0.32 ERBB2 2.00 1.33
8 ALK 1.32 0.48 BRCA1 1.20 0.42 ABL1 3.00 1.44
9 ABL1 1.37 0.83 BRCA2 1.20 0.42 AKT1 1.00 1.44
10 BCR 1.42 0.69 ESR1 1.20 0.63 AKT2 1.00 1.56
11 PTEN 1.53 0.51 ABL1 1.30 0.95 ALK 2.00 1.56
12 AKT1 1.58 0.77 IGF1R 1.40 0.52 PTEN 2.00 1.56
13 MET 1.58 0.61 MET 1.40 0.52 AKT3 2.00 1.67
14 AKT2 1.63 0.76 RET 1.40 0.52 CTNNB1 1.00 1.67
15 TP53 1.63 0.76 MAP2K1 1.50 0.53 BCR 3.00 1.78
16 AKT3 1.68 0.75 MAP2K2 1.50 0.53 CDKN2A 1.00 1.78
17 ESR1 1.68 1.11 PTEN 1.50 0.53 JAK2 1.00 1.78
18 IGF1R 1.68 0.82 RAF1 1.50 0.53 MET 1.00 1.78
19 RET 1.68 0.75 TP53 1.50 0.53 TP53 1.00 1.78
20 JAK2 1.74 0.87 BCL2 1.60 0.70 ATM 1.00 1.89
21 MAP2K1 1.74 0.73 KDR 1.60 0.52 HRAS 2.00 1.89
22 MAP2K2 1.74 0.73 PARP1 1.60 0.52 MLH1 2.00 1.89
23 PARP1 1.74 0.81 PARP2 1.60 0.52 MSH2 2.00 1.89
24 PARP2 1.74 0.81 PDGFRA 1.60 0.52 NRAS 2.00 1.89
25 BCL2 1.84 0.69 PDGFRB 1.60 0.52 PARP1 2.00 1.89
26 FLT1 1.89 0.88 AKT1 1.70 0.95 PARP2 2.00 1.89
27 FLT3 1.89 0.94 AKT2 1.70 0.95 DHFR 2.00 2.00
28 NRAS 1.89 0.94 AKT3 1.70 0.95 FLT3 1.00 2.00
29 CHEK1 1.95 0.71 FLT1 1.70 0.48 IGF1R 1.00 2.00
30 KDR 1.95 1.03 JAK2 1.70 0.67 MAP2K1 1.00 2.00
31 MSH2 1.95 0.85 AR 1.80 0.92 MAP2K2 1.00 2.00
32 PDGFRA 1.95 1.08 CHEK1 1.80 0.63 RET 2.00 2.00
33 PDGFRB 1.95 1.08 CHEK2 1.80 0.63 BCL2 3.00 2.11
34 PIK3CA 1.95 0.91 ERBB3 1.80 0.79 CDH1 1.00 2.11
35 CDKN2A 2.00 0.58 ERCC1 1.80 0.63 CHEK1 1.00 2.11
36 CHEK2 2.00 0.75 FGFR2 1.80 0.42 FLT1 1.00 2.11
37 MLH1 2.00 0.94 FLT3 1.80 0.42 MDM2 1.00 2.11
38 RAF1 2.00 0.88 PIK3CA 1.80 0.63 PIK3CA 1.00 2.11
39 RARA 2.00 0.88 TYMS 1.80 1.03 RARA 2.00 2.11
40 AR 2.05 1.18 VHL 1.80 0.92 SRC 1.00 2.11
41 FGFR1 2.05 0.91 ERBB4 1.90 0.74 AURKA 2.00 2.22
42 SRC 2.05 1.08 ESR2 1.90 0.99 AURKB 2.00 2.22
43 CTNNB1 2.11 0.81 FGFR1 1.90 0.57 CHEK2 1.00 2.22
44 DHFR 2.11 0.74 FGFR3 1.90 0.32 ESR1 4.00 2.22
45 ERBB3 2.11 0.94 NOTCH1 1.90 0.32 FGFR1 1.00 2.22
46 HRAS 2.11 0.88 NOTCH2 1.90 0.32 IDH1 4.00 2.22
47 MDM2 2.11 0.74 NOTCH3 1.90 0.32 IDH2 4.00 2.22
48 MGMT 2.11 1.05 NOTCH4 1.90 0.32 MGMT 2.00 2.22
49 ERBB4 2.16 0.90 NRAS 1.90 0.88 SMAD4 1.00 2.22
50 FGFR2 2.16 0.96 PIK3CB 1.90 0.57 AR 3.00 2.33

Abbreviation: SD, standard deviation.

ⴱ
As determined by a group of 19 experts, consisting of drug developers and genome scientists.

www.jco.org © 2012 by American Society of Clinical Oncology 15

Downloaded from jco.ascopubs.org on February 20, 2013. For personal use only. No other uses without permission.
Copyright © 2012 American Society of Clinical Oncology. All rights reserved.
 CORRECTIONS

Author Corrections
The October 10, 2005 editorial by Workman and Johnston, en- Patrick G. Johnston in the following two categories: “Employ-
titled, “Genomic Profiling of Cancer: What Next?” (J Clin Oncol ment or Leadership Position (Compensated)” and “Stock
23:7253-7256, 2005) contained an error. Ownership.”
In the Authors’ Disclosures of Potential Conflicts of Inter- The authors apologize for the mistake.
est section, Almac Diagnostics should have been disclosed for DOI: 10.1200/JCO.2012.42.7823; published April 1, 2012

■ ■ ■

The February 1, 2012 editorial by Paskett, entitled, “Cancer “Cancer Health Disparities: Moving From Why They Occur to
Heath Disparities: Moving From Why They Occur to How How They Can Be Prevented.”
They Can Be Prevented” (J Clin Oncol 30:354-356, 2012), con- The online version has been corrected in departure from
tained an error. the print.
The Title should have read: The author apologizes for the mistake.
DOI: 10.1200/JCO.2012.42.6379; published April 1, 2012

■ ■ ■

Journal Corrections
The February 20, 2012 editorial by Pfister, entitled, “Off-Label Mullins CD, Montgomery R, Abernethy AP, et al: Recom-
Use of Oncology Drugs: The Need for More Data and Then Some” mendations for Clinical Trials of Off-Label Drugs Used to Treat
(J Clin Oncol 30:584-586, 2012), contained an error. Advanced-Stage Cancer. J Clin Oncol 30:661-666, 2012.
The editorial was linked to an accompanying article on Journal of Clinical Oncology apologizes for the mistake.
page 647 of the issue, whereas it should have been the article on
page 661, as follows: DOI: 10.1200/JCO.2012.42.7831; published April 1, 2012

■ ■ ■

The February 20, 2012 special article by Tran et al, entitled, editorial on page 584 of the issue and should not have been
“Cancer Genomics: Technology, Discovery, and Translation” linked to any editorial.
(J Clin Oncol 30:647-660, 2012), contained an error. Journal of Clinical Oncology apologizes for the mistake.
The article was erroneously linked to an accompanying DOI: 10.1200/JCO.2012.42.7849; published April 1, 2012

■ ■ ■

© 2012 by American Society of Clinical Oncology 1149

Downloaded from jco.ascopubs.org on February 20, 2013. For personal use only. No other uses without permission.
Copyright © 2012 American Society of Clinical Oncology. All rights reserved.