Stanley R. Hamilton*
Division of Pathology and Laboratory Medicine, Department of Pathology, Unit 085, The University of Texas MD Anderson Cancer Center,
1515 Holcombe Blvd, Houston, TX 77030, USA
A R T I C L EI N F O
Received 24 February 2012
Accepted 27 February 2012
Available online 23 March 2012
A B S T R A C T
Molecular pathology as applied to neoplasia is a rapidly expanding component of the dis-
cipline of pathology that uses molecular biology tools in addition to conventional morpho-
logic, immunohistochemical and chemical analyses of abnormalities in tissues and cells to
understand the etiology and pathogenesis of tumors, establish their diagnosis, and contrib-
ute to prognostication and therapeutic decisions for cancer patient care. Biomarkers are
a fundamental component of personalized cancer care, and the discipline of molecular pa-
thology therefore contributes throughout the continuum from biomarker research to use in
standard-of-care personalized cancer therapy. This brief review addresses some of the spe-
cific roles of molecular pathology in that continuum.
ª 2012 Published by Elsevier B.V. on behalf of Federation of European Biochemical Societies.
1. Definitions and context
Personalized cancer medicine and other similar terms are now
widely used but have a large number of definitions and conno-
tations depending upon the context of use, ranging from the
generic to the highly specific. At the generic end of the spec-
trum, personalization is all-encompassing and is the tailoring
of all aspects of cancer care to the spectrum of characteristics
of each individual patient (President’s Council of Advisors on
Science and Technology, 2008). A far more specific definition
is the use of genetic biomarkers and/or pharmacogenomic
testing to tailor an individual’s preventative care or therapy
for cancer (Burke and Zmmern, 2004).
Personalized cancer therapy is influenced by the character-
istics of the patient’s tumor, genetic constitution, and environ-
mental exposures and lifestyle, as well as the characteristics of
the treatment modalities including surgery, radiation therapy,
and chemotherapy. In the case of chemotherapy, these charac-
teristics can be referred to as pharmacogenomics based on tu-
mor alterations, pharmacogenetics based on patient germline
and lifestyle, and pharmacokinetics/pharmacodynamics based
upon the characteristics of the therapeutic agents. Molecular
and the patient who has that tumor for use in clinical
Molecular pathology as applied to neoplasia is a rapidly
expanding component of the discipline of pathology that
uses molecular biology tools in addition to conventional mor-
phologic, immunohistochemical and chemical analyses of ab-
normalities in tissues and cells to understand the etiology and
pathogenesis of tumors, establish their diagnosis, and con-
tribute to prognostication and therapeutic decisions for can-
cer patient care. The discipline of molecular pathology thus
contributes throughout the continuum of biomarker research
leading to incorporation into standard-of-care personalized
cancer therapy. This continuum includes discovery; assess-
ment of viability and feasibility; planning; development, inte-
gration and verification; validation including clinical trials;
launch readiness and release; application in patient care,
and subsequent post-implementation evaluation of perfor-
mance (Phillips et al., 2006). Therapeutic agent development
also utilizes molecular pathology in research and clinical ap-
proaches to target identification, attempts to distinguish
5For special Molecular Oncology issue organized by Ulrik Ringborg, MD, PhD.
* Tel.: þ1 713 792 2040; fax: þ1 713 792 4094.
E-mail address: firstname.lastname@example.org
1574-7891/$ e see front matter ª 2012 Published by Elsevier B.V. on behalf of Federation of European Biochemical Societies.
available at www.sciencedirect.com
M O L E C U L A R O N C O L O G Y 6 (2012) 177e181
between driver alterations that are pathogenic and passenger
alterations that do not directly control tumor behavior, as-
sessment of the effects of agents on reputed targets and resul-
tant downstream responses in pathways, identification of
biomarkers for agent responseand resistance, and rational se-
lection of combinations of therapeutic agents.
2. Clinical biomarker development
The processes for bringing a biomarker into clinical use in-
volve many sequential and parallel steps. Initial data must
be obtained to support the investment of resources to develop
a biomarker. Definition of the intended clinical use is a key
first step (Brock, 2012), and the selection of the patient popula-
tions and specimens to match the intended clinical use is es-
sential to providing high-quality data. The biomarker assay
must be validated at the laboratory level and carefully quality
controlled includinguse of
proficiency-tested personnel (Holden et al., 2011) in order to
provide the basis for clinical validation. Study design is impor-
tant with sufficient sample size for statistical power and the
ability to demonstrate that a biomarker test improves clinical
outcome. The Institute of Medicine of the National Academies
in the United States released a comprehensive report with
recommendations on omics assays (Micheel et al., 2012).
3. Laboratory assay methodologies
The laboratory methodologies and technologies available for
use in molecular pathology are extensive and address a wide
variety of analytes including chromosomal DNA, mitochon-
drial DNA, messenger RNA, non-coding RNA that includes
McDermott et al., 2011). As the laboratory-based “-omics” in-
cluding genomics, transcriptomics, proteomics, metabolo-
mics, etc., have emerged, bioinformatics and biostatistics
have moved front and center in the analyses and interpreta-
tion of the resulting massive data sets (Vickers, 2008).
DNA evaluation includes a variety of sequencing ap-
proaches ranging from directed characterization of individual
genes through Sanger sequencing and pyrosequencing to
broad-scale “next generation” sequencing that can be applied
across all exomes or the whole genome (McDermott et al.,
2011). The availability of various methodologies for DNA se-
quencing provides the opportunity to employ a number of ap-
proaches to tumor characterization. These range from
targeted sequencing of specific individual genes (Sanger and
pyrosequencing) to whole genome sequencing that addresses
the breadth of somatic mutations in tumor DNA in compari-
son to germline sequences, and germline mutations and poly-
morphisms compared to “normal” sequences. Intermediate to
these methods is the evaluation of the transcribed portion of
the genome through whole exome sequencing. These meth-
odologies have advantages and disadvantages, with inverse
relationships between the amount of information that is
obtained and the ease of analysis.
Identification of the most effective sequencing strategies is
important as these technologies move into clinical utilization.
lipids(Louiset al., 2011;
Multiplex techniques can identify the whole range of muta-
tions in tumors and define changes in specific biologic path-
ways that have potential to be targeted by therapeutic
agents. A major challenge in clinical usage is the reporting
of results in a meaningful fashion (Louis et al.,2011) and deter-
mining the efficacy of agents that are selected on the basis of
mutation, including novel combinations of agents for which
toxicologic data and drug interaction information may not
These newer methodologies have led to the recognition of
the remarkable extent of nucleotide sequence variation in the
germlinegenomeand in the somaticabnormalities in cancers.
The methodological advances identified a small subset of can-
cers with hypermutation characterized by extremely frequent
sequence alterations as compared to germline sequence
(Stratton et al., 2009; Kumar et al., 2011). Sequencing can
also address copy number variation in the form of amplifica-
tions and deletions, as can older techniques such as fluores-
cent in situ hybridization (FISH) and analysis for loss of
heterozygosity in polymorphic areas of the genome. Rear-
rangements of chromosomal structure in the form of translo-
cations and inversions occur and can be detected by
cytogenetics and by sequencing. Methylation of CpG islands
in the promoters of many genes that suppresses transcription
and alter chromosomal structure can be analyzed by sequenc-
RNA evaluation addresses two major groups of analytes:
messenger RNA (mRNA) that is translated into proteins and
the much more prevalent non-coding RNAs that include
microRNAs (miRNAs) responsible for regulation of gene ex-
pression. Individual genes can be assessed by reverse tran-
scription into DNA and polymerase chain reaction for
quantitation (QRT-PCR), but the transcriptome can also be
evaluated by gene expression profiling with hybridization to
chips that are spotted with cDNA or by RNA sequencing.
Protein analysis has been a mainstay of molecular pathol-
ogy for decades through the application of immunohisto-
chemistry. Recent advances in proteomic methodologies
now permit broad and detailed unbiased characterization of
proteins, but antibody-based technology applied to reverse
phase protein arrays can also address simultaneously the al-
terations in large numbers of selected proteins, including
post-translational modifications such as phosphorylation. Re-
cent technical advances have led to the ability to assess lipids,
including lipidomic methodologies, and metabolomics can
characterize a wide variety of metabolic products.
The large number of available methodologies is daunting,
but available evidence on the importance of individual types
of alterations suggests that many different technologies will
be needed for clinical applications. For example, sequencing
of KRAS has become standard-of-care for patients with colo-
rectal cancer eligible for treatment with antibodies to epider-
mal growth factor, copy number variation evaluation of Her2
for breast cancer and trastuzumab, translocations involving
Bcr and Abl in chronic myelogenous leukemia for imatinib
and EML4 and Alk in lung cancer for crizotinab, methylation
of MGMT for temazolamide for glioblastoma multiforme,
and gene expression panels for recurrence score in breast can-
M O L E C U L A R O N C O L O G Y 6 (2012) 177e181
biomarkers become available in various tumor types, multi-
plex testing must expand for cost-effective evaluation of tu-
mor in patients with multiple therapeutic options.
4. Co-alterations and pathway dysregulation
The single greatest challenge for molecular pathology at pres-
ent is the integration of the many sources of data into mean-
ingful and actionable knowledge that can be applied to
improve patient outcomes. The complexity of the alterations
in individual cancer cells multiplied by the intratumoral het-
erogeneity among the population of cancer cells within a tu-
mor along with the temporal interactions of the cancer cells
with their changing microenvironment populated by stromal
and inflammatory cells and a plethora of molecules from var-
ious sources is truly staggering. This complexity has long been
apparent when individual methodologies were applied to mo-
lecular pathology, but the comprehensive approaches such as
The Cancer Genome Atlas and International Cancer Genome
Consortium (The Cancer Genome Atlas Research Network,
2008; Cancer Genome Atlas Research Network, 2011) have
pointed out the incredible frequency of co-alterations, the ex-
tent of dysregulation of numerous cellular pathways, and the
crosstalk among the pathways. Intertumoral heterogeneity is
obvious, and it is now glaringly apparent that no two patients
ever have molecularly identical tumors. The hope for clinical
application of molecular pathology is that the crucial alter-
ations in pathways, rather than the details of how each path-
way is altered, can be characterized eventually to identify
Decision-making on biomarker analysis
Criteria usable in decisions regarding the readiness of a bio-
marker assay for research and/or clinical use have not been
strength or levels of evidence for decision-making. Green and
Byar presented in 1984 an eight point scale for determining
treatment efficacy based on analysis of existing data sets.
Their categories ranged from anecdotal case reports to con-
firmed randomized controlled clinical trials (Green and Byar,
1984). Hayes et al. reported in 1996 the much more broadly ap-
plicable Tumor Marker Utility Grading System (TMUGS) that
addresses biomarker use in determining risk, screening, dif-
ferential diagnosis, predicting relapse or progression of pri-
mary or metastatic disease, predicting response to therapy,
and monitoring the course of disease to detect relapse or fol-
low detectable disease. The system included a utility score
and level of evidence number for process and endpoint asso-
ciationwith a biologic as well as with indicatorsofpatient out-
come (survival, disease-free survival, quality of life, and cost
of care). The comprehensive nature of the system is also its
deterrent to use due to the complexity. Lassere and colleagues
reported in 2007 a schema that is based on target, study de-
sign, statistical strength, and penalties (Lassere et al., 2007)
that has not gained wide acceptance. In the absence of quan-
titative approaches, opinion of experts on the breadth and
depth of evidence continue to guide most decision-making,
oftenthroughguidelines issuedby professional organizations.
Despite the weak evidence-based and haphazard ap-
proaches, numerous markers have moved into routine usage.
Recent literature reviews have summarized the current status
of clinical molecular testing for a variety of common cancers,
e.g. breast, colon, lung, pancreatic, and thyroid tumors, sarco-
mas, melanomas, and tumors of uncertain origin (Igbokwe
and Lopez-Terrada, 2011).
development and use
Pre-analytic challenges for biomarker
6.1. Complexity of molecular changes in neoplasms
As described above, the complexity of the molecular changes
in neoplasms is the most formidable challenge to biomarker
development. This situation is due to the large number of op-
portunities that are afforded, uncertainty about the drivers,
non-static temporal changes, reactive pathways, and intra-
and intertumoral heterogeneity. Molecular classifications of
tumors have potential to identify entities with clinical impor-
tance, if clinical associations can be identified and validated.
6.2. Specimen acquisition and qualification
An essential component of molecular pathology is the provi-
sion of high-quality tissue specimens for both research and
clinical care activities. Great emphasis has been placed on re-
search biorepositories with biospecimens for translational re-
search (National Cancer Institute, 2011). The importance of
population-based specimen acquisition and biospecimens
from patients enrolled in clinical trials has increased, as suc-
cessful studies with specimens from these two sources pro-
vide high levels and broad breadth of evidence for moving
biomarkers along the pathway to clinical application. The
methodologies for molecular pathology characterization put
new emphasis on frozenand appropriately stabilizedfixed tis-
sue, in addition to the routine formalin-fixed paraffin embed-
ded specimens generally available in pathology laboratories.
This new emphasis is necessitated by the analytic methods
that cannot address, or address poorly, even modestly de-
graded specimens. A change in laboratory approaches in all
pathology laboratories is needed to provide the high-quality
specimens, as buffers/fixatives for preservation of labile ana-
lytes (messenger RNA, proteins, phosphoproteins, etc.) are
not generally available. The ability to down-scale tissue quan-
tity is also an issue, as smaller specimens derived from inter-
ventional radiology procedures and endoscopic biopsies will
be the only appropriate tissue available in many efforts di-
rected at personalized cancer therapy. Amplification of some
analytes may increase the amount of material but may be un-
suitable due to perturbation from the native state. Qualifica-
tion of specimens for analysis of intact tumor is especially
problematical with destructive routine histopathology, and
needed to provide the maximum amount of available high-
M O L E C U L A R O N C O L O G Y 6 (2012) 177e181
Fit-for-purpose selection of tissue sources is important for
analytes of interest that vary in their characteristics. Elapsed
time from interruption of blood supply to stabilization of the
specimen is crucial for labile analytes. The much longer
time interval between specimen acquisition and clinical deci-
sion can also be important, as the decision to analyze primary
tumor, synchronous or metachronous metastatic tumor, or
recurrent tumor before or after prior therapy has the potential
to influence the utility of the laboratory analysis results. Inva-
sive procedures, often by interventional radiologists, are
needed to acquire tissue from common metastatic sites such
as the liver and lungs. There is therefore great interestin using
circulating tumor cells and nucleic acids because of their easy
access by phlebotomy, availability contemporaneous with
need for therapeutic decisions, and potential for sequential
analyses, but the representativeness of a patient’s tumor is
an issue that remains to be addressed.
Biorepositories have extensive responsibilities in tracking
consent, addressing sample collection requests, determining
the time of collection and processing, and maintaining inven-
tory control with location of specimens. In addition, sample
retrieval, checkout, and distribution must be managed.
Many biorepositories also annotate the specimens with clini-
cal information from the patients who provide the specimens.
Standard operating procedures for specimens, governance
documentation, access and prioritization processes, review
of applications for use, and administrative interactions are
needed. The National Cancer Institute’s Best Practices for Bio-
specimen Resources (National Cancer Institute, 2011) provides
a comprehensive reference, and other organizations have
guidelines and/or accrediting activities for clinical specimen
biorepositories in preparation or available.
6.3. Intratumoral heterogeneity: tumor and stroma
Although most cancers are clonal proliferations, evolution of
subpopulations (subclones)of tumor cellsiswell-knownto oc-
cur during tumor growth. The size and topographical distribu-
tion of the abnormal subpopulations in tumors are as yet
poorly studied. It is unclear if small subpopulations, that in
turn may be difficult to detect with some analytic methodolo-
gies, are important in therapeutic choices.
In addition to the characteristics of the tumor cell popula-
tion, the potential effects of stroma composed of non-
neoplastic cells must be addressed in specimens for bio-
inant of the tumor cells that complicates analyses by
contributing non-neoplastic material into the analysis mate-
rial, but it is now recognized that tumor microenvironment
is a key element. Analyses may need to address stromal cells
as well as tumor cells to achieve a complete understanding of
7. Biomarkers in clinical trials
An important goal in the application of biomarkers is to make
better use of currently available therapeutic agents by finding
subsets of responders, as well as contribute to the develop-
ment of new agents in clinical trials. The development of
predictive biomarkers associated with chemotherapeutic
and targeted agents is particularly difficult and is best done
in the settingof clinical trials with carefully selectedand mon-
itored patient cohorts. The clinical challenges include the low
response rates for many agents, the use of combination che-
motherapy that makes separation of biomarkers for individ-
ual agents difficult, differences in dosage and timing of drug
administration that may affect the informative timepoints
for biomarker assessment, and the range of biomarkers and
methodologies that are available. A practical problem with
combination therapy is the origin of agents from different
pharmaceutical and biotechnology companies with the asso-
ciated intellectual property issues. Biology is also important,
as sensitivity and resistance of tumor cells are complex bio-
logical processes. The mechanisms of effects of different ther-
apeutic agents influence this complexity, e.g. cytostatic as
compared to cytotoxic agents. Trafficking of signaling through
pathways is at an early stage of understanding, and the re-
sponse capabilities of tumors after perturbation by therapeu-
tic agents is not predictable from baseline status in many
instances. Plasticity of the cancer cell population, including
the acquisition of cancer stem cell characteristics after expo-
sure to agents, remains to be explained and addressed.
Resources must be in place for successful biomarker devel-
opment in clinical trials, and these include tissue repositories;
data management,bioinformatics and biostatistics;sourcesof
funding; and timing of biomarker development relative to
therapeutic agent development. Multiple assays must be
used for assignment of patients to arms of a trial with differ-
ent agents when each agent has an associated biomarker,
but in order to be cost-effective and conserve valuable tumor
specimens, multiplex assays (Roychowdhury et al., 2011) with
regulatory compliance are needed. Gene patents influence the
ability of clinical laboratories to develop assays, including the
needed multiplex assays, because patent rights can lead to re-
quirements for licensing or to submit all testing to a commer-
cial laboratory owned by the patent holder (Chen et al., 2010).
At present, the status of gene patents in the United States re-
mains to be determined by the Supreme Court.
The highest level of evidence for clinical use of a biomarker
comes from an integral marker or integrated marker adaptive
randomization clinical trial in whichthe biomarker results de-
termine assignment to a therapeutic arm (Hayes et al., 1996).
These challenging trials must meet regulatory compliance
for performance of the assay, and use of a central laboratory
has major advantages due to the concentration of expertise,
instrumentation, and economies of scale. As biomarkers be-
come established, distributed laboratories are far more com-
monly used. The source of funding to develop biomarkers is
often problematical due to contractual issues. Payers in clini-
cal trials include the trial sponsors, the National Cancer Insti-
tute, and third party payers, and appropriate distribution of
cost is often difficult.
8. Standard-of-care application of biomarkers
The approaches to appropriate use of biomarkers in routine
patient care are unsettled at present. On one hand, the level
of confidence in the performance of a biomarker for
M O L E C U L A R O N C O L O G Y 6 (2012) 177e181
therapeutics must be sufficiently high for physicians to decide
to give or withhold drugs from a particular patient. On the
other hand, the promises and cache of novel biomarkers de-
rived from state-of-the art science have great attraction for
their use. Direct-to-patient and direct-to-physician marketing
of unproven biomarkers is commonplace and often done in
the absence of peer-reviewed publication of laboratory valida-
tion of the methodologies, let along clinical validation of the
utility of the biomarkers in managing patients.
A variety of providers of molecular diagnostics is available,
including hospital laboratories, large and small reference lab-
oratories, and diagnostic assay companies. The current regu-
latory environment generally addresses the quality of the
laboratory analysis of biomarkers. In most countries, regula-
tory agencies have major impact on the utilization of molecu-
lar pathology in clinical applications. In the United States, the
Clinical Laboratory Improvement Amendments of 1988 (CLIA-
88) govern the accreditation of laboratories that provide labo-
ratory results for use in patient care. In addition, a wide vari-
ety of professional organizations have published guidelines
and recommendations for use of biomarkers, and various or-
ganizations have specific requirements governing the usage
of laboratory-developed tests (LDTs). As assays have become
increasingly complex, involving informatics algorithms as
well as multiplex testing, quality control/quality assurance
becomes an increasing challenge. In addition, the Food and
Drug Administration (FDA) in the United States has estab-
lished specific companion diagnostics based on specific man-
ufacturer’s assays that are linked to the approval of associated
therapeutic agents (Brock, 2012).
The circumstances around reimbursement for molecular
pathology in the clinical setting require further development.
The criteria and processes for determining a molecular diag-
nostic test is “standard-of-care” in a specific clinical setting
have not been established and are inconsistent among payers.
Development of documentation of medical necessity and bill-
ing compliance as well as post-implementation evaluation of
outcomes and clinical effectiveness remain poorly addressed.
9. Future directions
Molecular pathology will continue to grow in importance as
the continuum from biomarker research to clinical applica-
tions in personalized cancer therapy continues to expand
and become more robust. The broad areas for biomarker ap-
plication to risk identification, screening, surveillance, diag-
resistance, and monitoring for response and recurrence are
well-known, but the use-cases for specific biomarkers need
to be established. Ideally, future biomarker development
should be prioritized for greatest impact on improving patient
care with rapid laboratory method validation, clinical valida-
tion, and subsequent access for patients. Technological ad-
vances will continue to provide new methodologies to be
evaluated and applied in molecular pathology laboratories
that need to be resourced to acquire appropriate material
from patients, develop and provide assays, provide quality
control/quality assurance, and meet regulatory and reim-
R E F E R E N C E S
Brock, T.K., 2012. Diagnostic, prognostic, and predictive: the
dynamic role of molecular technology in personalized
medicine. Am. Lab., 21e25. January.
Burke, W., Zmmern, R.L., 2004. Ensuring the appropriate use of
genetic tests. Nat. Rev. Genet. 5, 955e959.
Cancer Genome Atlas Research Network, 2011. Integrated
genomic analysis of ovarian carcinoma. Nature 474,
Chen, Q., Ye, Z., Lin, S.-C., et al., 2010. Recent patents and
advances in genomic biomarker discovery for colorectal
cancers. Recent Patents on DNA & Gene Sequences 4, 86e93.
Micheel, C.M., Nass, S.J., Omenn, G.S. (Eds.), 2012. Evolution of
Translational Omics. Lessons Learned and the Path Forward.
Committee on the Review of Omics-Based Tests for Predicting
Patient Outcomes in Clinical Trials. Institute of Medicine of
the National Academies. The National Academies Press,
Washington, D.C. March 23, 2012. www.nap.edu.
Green, S.B., Byar, D.P., 1984. Using observational data from
registries to compare treatments: the fallacy of omnimetrics.
Stat. Med. 3, 361e370.
Hayes, D.F., et al., 1996. Tumor marker utility grading system:
a framework to evaluate clinical utility of tumor markers. J.
Natl. Cancer Inst. 88, 1456e1466.
Holden, M.J., Madej, R.M., Minor, P., et al., 2011. Harmonization
through reference materials, documentary standards and
proficiency testing. Expert Rev. Mol. Diagn. 11, 741e755.
Igbokwe, A., Lopez-Terrada, D.H., 2011. Molecular testing of solid
tumors. Arch. Patho; Lab. Med. 135, 67e82.
Kumar, A., et al., 2011. Exome sequencing identifies a spectrum of
mutation frequencies in advanced and lethal prostate cancers.
Proc. Natl. Acad. Sci. USA 108, 7087e17092.
Lassere, M.N., et al., 2007. Definitions and validation criteria for
biomarkers and surrogate endpoints: development and testing
of a quantitative hierarchical levels of evidence schema. J.
Rheumatol. 34, 607e615.
Louis, D.N., Virgin, H.W., Asa, S.L., 2011. Next-generation
pathology and laboratory medicine. Arch Pathol. Lab. Med.
McDermott, U.M., Downing, J.R., Stratton, M.R., 2011. Genomics
and the continuum of cancer care. N. Engl. J. Med. 364,
National Cancer Institute, 2011. Best Practices for Biospecimen
Phillips, K.A., Van Bebber, S., Issa, A.M., 2006. Diagnostics and
biomarker development: priming the pipeline. Nat. Rev. Drug
Discov. 5, 463e469.
President’s Council of Advisors on Science and Technology, 2008.
Priorities for Personalized Medicine, p. 13.
Roychowdhury, S., et al., 2011. Personalized oncology through
integrative high-throughput sequencing: a pilot study. Sci.
Translational Med. 3, 1e10.
Stratton, M.R., Campbell, P.J., Futreal, P.A., 2009. The cancer
genome. Nature 458, 719e724.
The Cancer Genome Atlas Research Network, 2008.
Comprehensive genomic characterization defines human
glioblastoma genes and core pathways. Nature 455,
Vickers, A.J., 2008. Decision analysis for the evaluation of
diagnostic tests, prediction models and molecular markers.
Am. Stat. 62, 314e320.
M O L E C U L A R O N C O L O G Y 6 (2012) 177e181