Standardized decision support in next generation sequencing reports of somatic cancer variants.

M O L E C U L A R O N C O L O G Y X X X ( 2 0 1 4 ) 1 e1 5

available at www.sciencedirect.com

ScienceDirect www.elsevier.com/locate/molonc

Review

Standardized decision support in next generation sequencing reports of somatic cancer variants Rodrigo Dienstmanna,*, Fei Donga, Darrell Borgerb, Dora Dias-Santagataa, Leif W. Ellisenb, Long P. Lea, A. John Iafratea a

Massachusetts General Hospital and Harvard Medical School, Molecular Pathology Lab, USA Massachusetts General Hospital Cancer Center and Harvard Medical School, 55 Fruit St GRJ, Boston, MA 02114, USA

b

A R T I C L E

I N F O

A B S T R A C T

Article history:

Of hundreds to thousands of somatic mutations that exist in each cancer genome, a large

Received 9 February 2014

number are unique and non-recurrent variants. Prioritizing genetic variants identified via

Received in revised form

next generation sequencing technologies remains a major challenge. Many such variants

18 March 2014

occur in tumor genes that have well-established biological and clinical relevance and are

Accepted 26 March 2014

putative targets of molecular therapy, however, most variants are still of unknown signif-

Available online -

icance. With large amounts of data being generated as high throughput sequencing assays enter the clinical realm, there is a growing need to better communicate relevant findings in

Keywords:

a timely manner while remaining cognizant of the potential consequences of misuse or

Cancer

overinterpretation of genomic information. Herein we describe a systematic framework

Genomics

for variant annotation and prioritization, and we propose a structured molecular pathology

Next-generation sequencing

report using standardized terminology in order to best inform oncology clinical practice.

Report

We hope that our experience developing a comprehensive knowledge database of

Variant annotation

emerging predictive markers matched to targeted therapies will help other institutions implement similar programs. ª 2014 Federation of European Biochemical Societies. Published by Elsevier B.V. All rights reserved.

1.

Introduction

With the advent of Next Generation Sequencing (NGS) or massively parallel sequencing technologies, we have the promise of a complete genetic description of patient tumors to optimally direct therapy. Clinical laboratories increasingly

view large cancer genes panels as a cost-effective d and tissue-saving d alternative to running a series of multiple single-gene companion tests. Tremendous amounts of genomic data are being generated, with hundreds to thousands of variants observed in the coding regions of an individual’s cancer genome, including somatic single

Abbreviations: CAP, College of American Pathologists; COSMIC, Catalog of Somatic Mutations in Cancer; EMR, electronic medical records; MAF, mutant allele fraction; NCCN, National Comprehensive Cancer Network; NGS, Next Generation Sequencing; SNP, single nucleotide polymorphisms; TCGA, The Cancer Genome Atlas. * Corresponding author. Present address: Sage Bionetworks, Fred Hutchinson Cancer Research Center, 1100 Fairview N, Seattle, WA 98109, USA. Tel.: þ1 206 667 4265. E-mail address: [email protected] (R. Dienstmann). 1574-7891/$ e see front matter ª 2014 Federation of European Biochemical Societies. Published by Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.molonc.2014.03.021

Please cite this article in press as: Dienstmann, R., et al., Standardized decision support in next generation sequencing reports of somatic cancer variants, Molecular Oncology (2014), http://dx.doi.org/10.1016/j.molonc.2014.03.021

2


nucleotide variants, insertions and deletions (indels), copy number alterations, rearrangements and germline susceptibility loci (Garraway and Lander, 2013; Vogelstein et al., 2013). Molecular pathologists and cancer genomicists face a particular challenge in the reporting of the cancer genome. Manually annotating each single variant in terms of clinical significance in every possible tumor type is a daunting challenge. The large amount of data generated by high throughput assays and strain on the turnaround time drive the need for prioritization strategies for the identification and reporting of clinically significant genetic variants. Routine testing of full gene sequences as opposed to hotspots (Dias-Santagata et al., 2010) frequently identifies mutations of low frequency and unknown functional consequences, most of which are likely to be neutral or passenger alterations. On the other hand, some of these rare variants occur in cancer genes that have well-established clinical utility, driving tumorigenesis and tumor progression. The available scientific knowledge on these mutations should be presented in the report, so that physicians and patients can make evidence-based decisions in a responsible fashion. In addition, the availability of genetic results may provide a strong rationale for treatment with matched targeted agents in clinical trials, with the potential of directly benefitting the patient and accelerating the drug development process (Dienstmann et al., 2013). Consolidating so much information into a very discrete report that clearly identifies the clinical significance while preserving observations that can be further looked into by the clinician is not an easy undertaking. As physicians trained in fields other than genetics are playing a more central role in the ordering and reviewing of genetic test results, the importance of translating genomic data into informative reports is further increased. In cancer genomics, performing NGS in the clinical laboratory is a multistep process that typically involves sample acquisition and quality control, DNA extraction, library preparation, sequencing and genomic data generation. As illustrated in Figure 1, the process continues with three dynamic pipelines for data analysis: (i) bioinformatics tools for variant identification; (ii) variant annotation and prioritization; and (iii) interpretation of clinical significance and reporting to clinicians (Van Allen et al., 2013; Watt et al., 2013). In this manuscript we discuss the challenges involved in the variant annotation and prioritization process and describe how genomic data can be translated into structured evidence-based reports. We hope that our experience developing the framework for clinical interpretation of somatic cancer variants and a comprehensive knowledge database of emerging predictive markers matched to targeted therapies will help other institutions implement similar programs.

2.

Variant annotation and prioritization

The report generation process starts with standardized definitions by the molecular pathology laboratory of “reportable” and “not reportable” variants. Following variant identification using bioinformatics pipelines, a computational engine is

needed in order to parse the variants and suppress those that are irrelevant, highlight the ones which need manual curation and identify pertinent “wild-types” in each tumor sample. As discussed below, several annotation and prioritization parameters are taken into consideration so as to provide a stronger estimation of the functional significance of unknown and novel mutations. Useful tools include sequencing metrics variables, external germline single nucleotide polymorphisms (SNPs) and cancer databases for comparison of variants across populations, as well as prediction models for defining damaging/deleterious or potentially driver mutations.

2.1.

Computational/bioinformatics tools

When analyzing large cancer gene panels based on exome or whole genome sequencing, pairwise comparison with germline DNA plays a pivotal role. Subtracting the genetic variation of a non-cancerous “normal” genome from its cancerous counterpart allows the identification of the somatic mutations. Eliminating known harmless variants that are present in public (dbSNP) or in-house polymorphism databases and published studies such as the 1000 Genomes Project (Abecasis et al., 2012) and the Exome Sequencing Project (ESP6500)(Fu et al., 2013) is a very helpful strategy for reducing the candidate list of deleterious mutations. Additionally, different bioinformatic adjustments can be used in order to improve variant detection and deal with library preparation or sequencing artifacts along with sample characteristics, including tumor purity and heterogeneity. Adequate coverage in target regions needs to be assured not only for variant detection but also to accurately define pertinent “wild-types” in specific tumor types. The next step involves prioritizing missense, nonsense or splice-site mutations over synonymous and intronic variants. In order to consider the variant as real and reportable, it is also advised to establish a minimum threshold of mutant allele fraction (MAF), the number of alternate reads at the genomic position divided by the total number of reads e coverage e at the same site. This threshold should take into consideration tumor cellularity and also clinical context, as rare resistant subclones in the treatment-refractory setting might be of relevance. Therefore, known gene variants previously clinically annotated are generally prioritized irrespective of MAF. If available, comparison with gene expression data (RNA sequencing) of the same sample can help determine the possible functional effects of a mutation, as variants not transcribed are less likely driver genomic events and therefore are not prioritized in the annotation process. A fundamental aspect of the bioinformatics pipeline includes dealing with the mutational heterogeneity across the genome of a particular tumor, across different regions of the same tumor, across patients in a given tumor type, and across multiple tumor types. Sophisticated algorithms that incorporate DNA replication timing and transcriptional activity in the mutation call pipeline, such as MutSigCV, are able to identify true driver genomic events with higher accuracy (Lawrence et al., 2014, 2013). These models should be taken into consideration when implementing computational genomic methods in clinical molecular pathology labs.



3

Figure 1 e Variant analysis flowchart of next generation sequencing tests performed in clinical laboratories 1 KG: 1000 Genomes Project; AACR: American Association for Cancer Research; ASCO: American Society of Clinical Oncology; COSMIC: Catalog of Somatic Mutations in Cancer; dbSNP: database of single nucleotide polymorphisms; ESMO: European Society for Medical Oncology; ESP6500: Exome Sequencing Project; PharmGKB, Pharmacogenomics Knowledge Base TCGA: The Cancer Genome Atlas.


4

2.2.


Cancer variants databases

The most useful annotation tool for somatic variant interpretation involves the assessment of published cancer databases. Computational engines should directly link genetic variants to the Cancer Gene Census (http://cancer.sanger.ac.uk/cancergenome/projects/census/) or similar catalogs of genes for which mutations have been causally implicated in cancer (Lawrence et al., 2014), as well as the Catalog of Somatic Mutations in Cancer (COSMIC; http://cancer.sanger.ac.uk/cancergenome/projects/cosmic/) and The Cancer Genome Atlas (TCGA; http://cancergenome.nih.gov/, http://www.cbioportal.org/), large cancer databases that present prevalence of gene variants in different tumor types. Assessing whether a newly discovered alteration may be functionally relevant rests heavily on how many times it has been reported in these international database of mutations associated with cancer. Curated information on each event in cancer genomics research studies supports downstream clinical interpretation.

example is the Cancer-specific High-throughput Annotation of Somatic Mutations (CHASM) tool, specifically designed to distinguish driver from passenger somatic missense variants. It is trained on a positive class of drivers curated from the COSMIC database and a negative class of passengers variants generated in silico based on background base substitution in specific tumor types (Wong et al., 2011). Limitations include reduced coverage as compared to traditional algorithms e restriction to missense mutations e and the understanding that driver and passenger mutations are tumor type and contextdependent, possibly changing roles during cancer evolution and therapy (Zhang et al., 2013). Whether cancer-trained methods outperform more general predictors still needs further investigation. Recent studies suggest that no method or combination of methods exceeds w80% accuracy (Gnad et al., 2013; Gonzalez-Perez and Lopez-Bigas, 2011), indicating there is still significant room for improvement in functional prediction, such as development of specific algorithms for different classes of mutations.

2.3.

2.4.

Functional prediction models

Prediction of the putative functional effect of a mutation is a common problem already addressed in the context of studies of associations using germline SNPs, and several popular tools have been used for this purpose. These models annotate variants specifically with respect to evolutionary conservation, biochemical deleteriousness and functional importance scores, thereby facilitating the differentiation between functional and non-functional variants (Cooper and Shendure, 2011; Frousios et al., 2013; Zhang et al., 2014). At present, for alleles without prior functional analysis in genes that have been related to human cancer, such as nonhotspot/novel variants in known oncogenes and tumor suppressor genes, prediction algorithms based on evolutionary conservation patterns are often used. Sorting Intolerant from Tolerant (SIFT)(Kumar et al., 2009) and MutationAssessor (Reva et al., 2011), exploit the fact that sequences observed among living organisms are those that have not been removed by natural selection and sites with fewer observed substitutions are inferred to be under tighter constraints, having more deleterious effects when mutated. On the other hand, mutations in non-conserved residues are likely neutral. Other resources for predicting the effects of protein-coding sequence changes typically exploit the physico-chemical properties of amino acids and information about the role of amino acid side chains in protein structure. These in silico protein-sequence-based algorithms, such as PolyPhen2 (Adzhubei et al., 2010) are capable of leveraging both evolutionary and biochemical information. Despite having high sensitivity for the detection of damaging variants, prediction tools that rely on conservation and structure should be used with caution. In addition to the low specificity, these methods generally have limited value in annotating gain-of-function or switch-of-function mutations (Flanagan et al., 2010). Furthermore, most of these algorithms have been designed for research purposes with germline variants, and very few databases present clinically-oriented molecular annotation. Machine-learning scoring methods attempt to increase the predictive precision of somatic mutations in cancer. One

Rules for variant prioritization

Complex criteria involving multiple fields from different annotation sources are regularly used in order to select or filter out variants, as shown in Figure 1. Part of this process can be automated, although most of the work still needs to be done manually. The most valuable tool consists in leveraging the cancer literature, either generated in-house or derived from publicly available databases, for the prevalence of the variant across subsets of samples. Consequently, the genomic management software needs to be dynamic in nature, recognizing driver cancer mutations that have been previously annotated and reported. Additional tumor-specific variants with very low allele fractions and those considered silent mutations are typically excluded from further clinical interpretation. Novel variants in genes that have been causally implicated in cancer are prioritized when functional models predict damaging/deleterious scores, the alteration is in the phosphorylation loop of an oncogenic kinase or it alters the reading frame of a tumor suppressor gene.

3.

Report generation

After narrowing down the list of candidate variants, the biggest challenge is to interpret the remaining genomic alterations within a biological context. Potentially “reportable” variants can be grouped in three categories: (i) those that may have a direct impact on patient care and are considered “actionable”; (ii) those that may have “biological relevance” but are not clearly actionable; and (iii) those that are of “unknown significance”. Different groups have varying definitions for clinically “actionable”. This category can be restricted to variants matched to drugs that have been approved by regulatory agencies for the tumor that is being studied, but may also include those directing to off-label use of approved drugs, as well as variants that are matched to drugs being investigated in clinical trials. Academic laboratories recognize the fact that expanding the definition of “actionable” may increase challenges in clinical decision-



making as the results sometimes lead to regulatory issues regarding the use of targeted drugs in unapproved indications. While there have been a number of published reports showing that sequencing results are associated with dramatic responses to off-label use of a targeted agent (Iyer et al., 2012; Sen et al., 2012; Serra et al., 2013), the success rate of this approach is still unclear. In this context, patients might be referred to early clinical trials with matched experimental drugs, but this strategy is not widely available outside large institutions with drug development units. We believe that the most inclusive definition of an actionable mutation should be used, which accounts for variants that support treatment recommendation, enrolment in a particular clinical trial or have prognostic or diagnostic implications. Even though some institutions might decide not to report back to clinicians the variants of unknown clinical relevance, restricting the access only to variants with well-established clinical effects is not acceptable in academic centers with drug development and research programs. Likewise, reporting all variants in an uncategorized format and allowing clinicians complete control over what actions to take can be confusing and detrimental to patients. Keeping this in mind, actionable mutations should be fully curated. This requires a team of specialists with strong background in cancer biology, access to up to date information, careful evaluation of the published literature regarding the clinical significance of a mutation on a particular tumor type and the evidence of utility of matched targeted agents. To facilitate the interpretation of individual cancer data, several groups have implemented expert panels such as the “Sequencing Tumor Boards” or “Molecular Rounds” with up to 15 faculty members that share expertise in cancer genomics, bioinformatics, pathology, clinical genetics, bioethics and clinical oncology as well as experimental therapeutics. This strategy guarantees that decisions do not rest with individual doctors, but gathering specialists that are distributed across several departments or institutes in one location on a regular basis can be very challenging. As previously described for the variant annotation and prioritization engine, computational tools are needed in order to support rigorous analysis and clinical interpretation of comprehensive genomic data. Creating a workstation that is integrated to medical knowledge-based publicly available databases is crucial, so that trained molecular pathologists or cancer genomicists can analyze and generate reports of NGS-based molecular diagnostic studies. A user-friendly interface with search functions linked directly to highquality knowledge resources can assist the expert panel in obtaining clinically relevant information about a variant. It is a time consuming and complex task, considering that not many mutations have been validated with a high enough level of evidence to predict for response to targeted treatment. Very few specific mutations are listed in consensus management and reporting guidelines such as the National Comprehensive Cancer Network (NCCN) or College of American Pathologists (CAP), for example. Surprisingly, response data according to specific gene variants is missing in many published clinical trials and preclinical studies that evaluate predictive associations.

5

It is important to emphasize that for complete annotation of predictive cancer genomic alterations, information on four different layers is desirable: (i) gene; (ii) specific variant; (iii) drug or class-of-agent sensitivity/resistance patterns; and (iv) tumor type context. Most knowledge resources currently available cover information at limited levels: some focus on geneetumor associations, others only on geneedrug or drugetarget relationships. In addition, databases originally developed to enable preclinical research or annotate germline variants are of limited applicability for clinical oncology curation. The Therapeutic Targets Database (http://bidd.nus.edu.sg/group/cjttd/), DrugBank (http://www.drugbank.ca/), the Pharmacogenomics Knowledge Base (https://www.pharmgkb.org/) and the Drug Gene Interaction database (http://dgidb.genome.wustl.edu/), for example, are useful first-round curation resources, but shortfalls include the limited evidence of some gene-drug associations, lack of annotations at the gene variant level or pathway-wise analysis of possible downstream targets, and poor updating with current investigational agents (Griffith et al., 2013).

3.1.

Web-based tools and literature search

Clinical interpretation of most variants identified in NGSbased cancer diagnostic tests usually involves manual searching of the published literature, a burdensome process for those collating and making sense of the information. Bioinformatics systems for managing individualized cancer genomics data in the context of various alterations and drugs are being developed. The first step is mining on public knowledge databases with curated annotations on predictive, prognostic or diagnostic variants in cancer, such as My Cancer Genome (http:// www.mycancergenome.org/) or similar institution-specific databases such as Targeted Cancer Care (http://www.targetedcancercare.org/). These websites are the result of large institutional or national efforts to provide state-of-the-art information on cancer types, aberrant genes and variants within those genes that are targeted by approved or experimental therapies. Of note, information is linked to clinical trial databases, guiding patients and physicians to genomicallydriven drug development programs. My Cancer Genome has an additional component called “DNA-mutation Inventory to Refine and Enhance Cancer Treatment”, or DIRECT database. This section is designed to support users studying very rare cancer mutations that only occur in a few cases and have much less supporting scientific evidence for their clinical significance. DIRECT is being used to catalog clinically relevant somatic mutations in the EGFR gene in non-small cell lung cancer patients and their response to EGFR tyrosine kinase inhibitors based on the detailed analysis of hundreds of published manuscripts (Yeh et al., 2013). The plan is to expand and incorporate data on all known mutations that have potential clinical significance in various types of cancer. A substantial body of evidence can also be retrieved in sources such as PubMed, Google Scholar and in abstracts or presentations from prominent cancer research conferences. The use of standardized search terms that include gene variant and tumor type followed by key terms such as “predictive”, “targeted therapy”, “response”, “inhibitor” or “prognosis” typically generates a list of publications with


6


information derived from clinical trials and preclinical research that has to be carefully reviewed.

3.2.

Interpretative challenges

Molecular pathologists and cancer genomicists facing NGS data have to deal with many potential interpretation challenges, as summarized and exemplified in Table 1 (Ascierto et al., 2013; Bose et al., 2013; Carpten et al., 2007; Carver et al., 2011; Chen et al., 2012; Cortes et al., 2012; Dahlman et al.,

2012; Engelman et al., 2007; Fong et al., 2009; Gartside et al., 2009; Gozgit et al., 2012; Guagnano et al., 2012; Heinrich et al., 2003; Hofmann et al., 2012; Jo et al., 2011; Prahallad et al., 2012; Sen et al., 2012; Sequist et al., 2013; Shi et al., 2012; Stone et al., 2012; Wagle et al., 2011; Whittaker et al., 2013). The knowledge on mutations in tumor specific contexts is prioritized, but curation of data derived from other tumor types can also give valuable information to clinicians. Considering the lack of reproducibility of many preclinical experiments, we believe that data derived from these studies

Table 1 e Challenges for interpreting genomic data in cancer. References Is this an activating or inactivating mutation?

1. Some mutations are activating and confer oncogene-addiction in specific contexts (FGFR2S252W or FGFR2N549K in endometrial cancer, effectively targeted by FGFR inhibitors ponatinib and BGJ398 in preclinical models). Others generate markedly reduced kinase activity and loss-of-function (FGFR2R251Q and FGFR2I648T in melanoma, with no predicted benefit with FGFR targeting).

(Gartside et al., 2009; Gozgit et al., 2012; Guagnano et al., 2012)

Does this mutation engender sensitivity to specific targeted therapeutics?

1. Some activating mutations based on in vitro models do not confer sensitivity to agents targeting the mutant protein (AKTE17K in breast cancer, effectively targeted by non-allosteric AKT inhibitors but not by allosteric AKT inhibitors; MEKC121S in melanoma, not inhibited by allosteric MEK inhibitors). 2. With inhibitors of the mutant kinase, level of sensitivity depends on the potency of the agent (EGFRT790M in lung cancer, resistant to first-generation EGFR inhibitors but higher sensitivity to novel irreversible inhibitors;ABL1T315I in chronic myeloid leukemia, resistant to imatinib and sensitive to ponatinib; ERBB2L755S in breast cancer, resistant to lapatinib and sensitive to neratinib). 3. With downstream pathway inhibitors, level of sensitivity strongly depends on the functional effects of the mutation (activating BRAFL597 and NRASQ61mutations confer sensitivity to MEK inhibitors in melanoma; BRAFY472C reduces kinase activity and confers sensitivity to dasatinib in lung cancer, but not to MEK inhibitors). 4. For rare variants in oncogenes, there is no definitive preclinical or clinical data to suggest sensitivity or resistance to targeted therapy (activating PDGFR mutations in the tyrosine kinase domain render tumors susceptible to PDGFR inhibitors in gastrointestinal stromal tumors, but the sensitivity of novel variants in the transmembrane domain of the gene is currently unknown). 5. Mutations that predict responsiveness to a therapy in some contexts, such as BRAF inhibitors in BRAFV600E mutant melanoma, may be associated with entirely different clinical interpretations in others. In colorectal cancer, for example, BRAFV600E mutations would direct towards combination of targeted therapies (BRAF plus EGFR inhibitors). It is important to emphasize that sensitivity is tumor context-specific and is influenced by concomitant genomic alterations.

(Carpten et al., 2007; Jo et al., 2011; Wagle et al., 2011)

(Bose et al., 2013; Cortes et al., 2012; Sequist et al., 2013)

(Ascierto et al., 2013; Dahlman et al., 2012; Sen et al., 2012)

(Heinrich et al., 2003)

(Prahallad et al., 2012)

How to select therapy in case of multiple genomic alterations?

1. The finding of concomitant genomic alterations in a tumor sample collected before systemic therapy can have important biological and therapeutic implications (in the setting of BRAFV600E mutant melanoma, NF1 loss predicts resistance to single-agent BRAF inhibitors; identification of BRAFV600Eand MEKP124S in melanoma does not predict resistance to BRAF inhibitors; in KRAS mutant lung cancer, loss-of-function STK11 mutations engender resistance to the combination of MEK inhibitors and docetaxel). 2. In relapsed samples after targeted therapies, identification of “acquired” genomic alterations may be linked to resistance mechanisms and help define subsequent therapies (in EGFR-mutant lung cancer progressing to erlotinib, the finding of MET amplification directs to combination therapy with EGFR and MET blockade). 3. On the other hand, prioritizing therapy is the setting of multiple “targetable” alterations is not as straightforward as in previous examples, especially when both targeted drugs are still in early phases of clinical development (activating KRASG12D mutation in pancreatic cancer, targeted by downstream PI3K pathway þ MEK inhibitors, often coexist with CDKN2A loss-of-function mutations, which may theoretically predict sensitivity to CDK inhibitors; in prostate cancer, PTEN deletion directs to PI3K pathway inhibitors and the coexistence of BRCA2 loss supports the use of PARP inhibitors).

(Chen et al., 2012; Shi et al., 2012; Wagle et al., 2011; Whittaker et al., 2013)

(Engelman et al., 2007)

(Carver et al., 2011; Fong et al., 2009; Hofmann et al., 2012; Stone et al., 2012)



should be interpreted and reported only when clinical validation is underway. As an example, mining cancer cell line encyclopedias of drug sensitivity and trying to infer potential predictive associations is typically uninformative to clinicians.

3.3.

Formatting the report

Previous studies evaluating single gene reports have suggested that patient care may be compromised as a consequence of poor communication between laboratories and clinicians (Lubin et al., 2008). Developing a framework to report NGS data to clinicians is complicated. In order to create accessible and content-rich reports, we took into consideration prior discussions with medical oncologists, pathologists, genetic counselors, bioinformaticians and laboratory directors. It became clear that physicians rely heavily on limited subsets of data and frequently apply rules based on strength of evidence in the decision-making process. It also helped us to identify where potential misunderstanding in the report could occur. Using this feedback, we realized that optimizing data visualization would be a critical aspect of the report generation. The traditional “narrative” style reporting is too cumbersome for the amount of data generated by large cancer gene panels, whole exome and whole genome sequencing. Medical oncologists prefer structured reports with results displayed in a more objective/straightforward manner rather than detailed descriptions of each genomic alteration. Consequently, webenabled technologies are a good alternative to text reports as they enable dynamic and interactive display of the NGS results, which could be accessed by providers and potentially patients in different formats. Embedding links to internal and external databases allows members of the team to further explore the results and the evidence used to guide the interpretation, including more detailed information on the gene, the variant, the drug or clinical trial matched to a particular genomic alteration and tumor type, as well as records of PubMed identification numbers for relevant clinical literature. Unfortunately, most laboratory information systems and electronic medical records (EMR) to date do not support data formatting and metadata (data associated with the result). Therefore, reports may need to be oversimplified to a static format for inclusion in the EMR. Wagle et al. reported the first framework to segregate genetic alterations derived from NGS tests on the basis of their predicted clinical utility (Wagle et al., 2012). The actionable category includes variants that predict tumor sensitivity or resistance to approved (tier 1) or experimental therapies (tier 2). These mutational categories are elegantly organized based on the strength of evidence supporting its predictive power. Each gene variantedrug association is prioritized as “clinically-validated in the same tumor type”, “limited evidence”, “evidence in another tumor type”, “preclinical evidence” or “theoretical”. Another category contains prognostic or diagnostic variants, either “clinically validated” or with “limited evidence”. Based on this comprehensive framework of tiers combined with levels of evidence we developed a simplified geneoriented approach to report somatic variants identified in NGS studies with the purpose of facilitating clinical decision-making, as shown in Figure 2. Our report provides the information in a hierarchical/categorical format, and

7

results are structured in tabular view. The content is formatted in such a way as to draw the clinician’s attention to associations with the highest level of evidence. Figure 3 illustrates the first page of the two different reports. All actionable - predictive, prognostic and diagnostic markers e are displayed first, followed by biologically relevant gene variants that warrant detailed annotation and pertinent negatives in the tumor being tested. All other variants that do not fall into the prior categories are displayed by pathway, as summarized in Table 2. Detailed information on matched genomically-driven clinical trials actively recruiting patients at our institution is presented in the last page of the report.

3.4.

Reporting predictive associations

Returning the results of predictive markers was a priority for clinical reporting, based on the principles of beneficence and respect for patients. Specific challenges regarding how to best communicate the strength of evidence for clinical decisionmaking had to be addressed, mainly because we decided to avoid writing long sentences with detailed information for these relationships in the report. We defined consensus associations as those: (i) linked to drugs approved or rejected by regulatory agencies in the context of a specific gene variant and tumor type; or (ii) those described in national guidelines as predicting response or resistance to specific therapies. Emerging predictive associations were classified in a hierarchical way based on the strength of evidence: (i) late trials, including evidence derived from trials that prospectively recruited patients based on genomic profiling as well as large trials with robust data suggesting sensitivity/resistance to targeted therapies based on retrospective analysis of biomarkers; (ii) early trials, referring to phase 1 or 2 studies with genomically-selected patients that show preliminary signs of efficacy (or lack of efficacy); (iii) case reports of dramatic responses to targeted therapies in a specific genomic context; and (iv) strong preclinical data that is being explored in clinical trials. In order to give unbiased and more detailed information with regard to the magnitude of the biomarker-related drug effects, we expanded the typical binary classification of responsive and resistant. In the advanced clinical setting, effect is outlined as responsive, resistant or not responsive (when an expected responsive effect is not observed). In preclinical models, biomarkeredrug associations are graded as sensitive, reduced sensitivity or resistant. Examples are seen in Figure 3 and Table 3. When predictive associations with agents approved for any indication are identified, the generic name of the drug is presented. In the case of investigational agents, the mechanism of action is displayed in the report.

3.5.

Reporting prognostic and diagnostic associations

We identified prognostic and diagnostic genomic alterations as particularly important in some disease settings, sometimes leading to changes in follow-up and/or treatment strategies. On the other hand, medical oncologists expressed concern about reporting detailed information on prognostic associations of genomic markers in cancer. First, the literature is full of inconsistent and even opposing results based on retrospective studies. Discrepancies in the prognostic impact of


8


Figure 2 e Somatic variant classification system and corresponding action-items.

TP53 mutations in non-small cell lung cancer, for example, illustrate the problem (Mitsudomi et al., 2000; Mogi and Kuwano, 2011; Scoccianti et al., 2012). Second, as patients have full access to the report, bad prognostic associations could lead to misinterpretation and anxiety, emphasizing the idea this information should be discussed in person taking into consideration additional clinical parameters. With this feedback, we decided to limit the report to prognostic markers with well-established associations in the same tumor type, with no description of the related outcome information. Common diagnostic associations are also described, mainly those favoring a specific tumor subtype, as exemplified in Table 3.

3.6.

Reporting other variants with biological relevance

Many variants in well-known cancer genes do not fall into the prior categories but still might be causally associated with the malignant phenotype. They can be grouped in three major categories: (i) variants in known cancer oncogenes; (ii) variants in established tumor suppressor genes; and (iii) other genomic alterations clearly implicated in tumor biology. Their relevance is justified by the anticipated pathway activation/

inactivation and “theoretical” actionability, when agents potentially targeting novel activating mutations in oncogenes or the downstream effects of loss-of-function mutations in tumor suppressor genes are available in the clinical setting. Therefore, the expected effect of the variant on protein function (gain- or loss-of-function) is also presented in the report, as it might give insights to the ordering physician with regards to therapeutic interventions in the investigational setting. Nevertheless, until functionality is validated by preclinical studies, it is appropriate to report these variants as nonactionable. Variants in genes that may function as oncogenes or tumor suppressor genes depending on the tumor context, such as NOTCH1 and FGFR2, are also reported in this section. Examples can be seen in Figure 3 and Table 3.

3.7.

Reporting pertinent negatives variants

Genes that have clear predictive, prognostic or diagnostic associations in each tumor type and unequivocally defined as “wild-type” are displayed in this section. Those not assessed by the assay are also presented in the report, as exemplified in Figure 3.



3.7.1.

Reporting variants of unclear significance

All other “reportable” uncharacterized gene variants can be displayed at the final section of the report, with no related annotations. Depending on the size of the gene panel, laboratories might choose not to return these results to clinicians. We believe that this information might be of interest in research institutions. The accelerated pace of

9

advances in the understanding of cancer genomics further emphasizes the importance of displaying all “reportable” variants in the report, which may become useful biomarkers in the near future. The most practical approach to handle variants of unknown biological/clinical significance is to present them according to the main pathway affected by the alteration, as shown in Table 2. Key geneepathways

Figure 3 e Illustrative examples of sequencing results describing somatic cancer variants with structured evidence-based classification. Please cite this article in press as: Dienstmann, R., et al., Standardized decision support in next generation sequencing reports of somatic cancer variants, Molecular Oncology (2014), http://dx.doi.org/10.1016/j.molonc.2014.03.021

10


Figure 3 e (continued)

associations are increasingly being reported in the literature (Garraway and Lander, 2013; Vogelstein et al., 2013). In some tumor types, such as renal cell carcinomas, mutations in genes involved in histone modification/chromatin remodeling might dominate a report, warning the medical oncologist-translational researcher about the importance of dysregulation in this pathway during cancer progression.

3.8.

Matching genomic alterations to clinical trials

Ideally, reports of NGS tests in oncology should include a list of clinical trials recruiting patients with specific genomic aberrations identified in the individual tumor sample. Our report presents matched investigational therapies available on-site, and the plan is to expand to multi-institutional databases in the near future. A current limitation for matching a patient’s tumor genotype to clinical trials is the lack of molecular annotations in notices of national registries, such as the U.S. National Cancer Institute clinical trials locator (www.clinicalTrials.gov). As an example, the search term “PIK3R1” does not identify any

matched trial, even though many PI3K pathway inhibitors in clinical development have a clear rationale for testing in tumors that harbor PIK3R1 activating mutations. There is an urgent need to address this shortfall.

3.9.

Reporting germline variants

The American College of Medical Genetics and Genomics (ACMG) recently published a minimum list of genes that should be reported to the patient when an incidental germline mutation associated with heritable risk of cancer or other diseases is identified and confirmed (Green et al., 2013). The group prioritized disorders where preventive measures and/or treatments were available and those in which individuals with pathogenic mutations might be asymptomatic for long periods of time. Only pathogenic mutations should be reported, considering the challenges of interpreting variants of unknown significance as incidental findings. Notably, the group acknowledged the fact that insufficient data on penetrance and clinical utility supports these recommendations. Considerable personnel


11


Table 2 e Variants of unknown significance are presented according to major pathways affected. Receptor tyrosine kinase signaling PI3K pathway RAS pathway MAPK/Stress kinase pathway JAK/STAT pathway TGF-beta signaling Wnt/beta-catenin signaling NF-KappaB signaling Notch signaling Hedgehog signaling DNA damage control/genome integrity Cell cycle/apoptosis Transcription factor/regulation Chromatin histone modification/SWI/SNF complex Epigenetics/DNA methylation Metabolism Protein homeostasis/ubiquitination Splicing Angiogenesis Immune system

contact the physician that ordered the test to discuss the findings of likely pathogenic germline variants. Variants with high MAFs are prioritized but we have not defined a cutoff for cancer genetics consultation. The medical oncologist is responsible for coordinating genetics specialist counseling and confirmatory tests. We hope this process to be streamlined in the near future and eventually, germline events become part of the full molecular pathology report, considering their unquestionable relevance for patients and their relatives.

3.10.

resources, including genetic counselors with specialized training, may be needed to ensure that patients understand the potential benefits and risks of receiving somatic and germline data and to support physicians in conveying such information. At present, our strategy is to directly

Reporting performance characteristics of the test

Specific regions interrogated by the assay and the coverage metrics by sample and target e including median depth, uniformity and percentage of target covered at the minimum level e should be described in every NGS assay, regardless of the application or platform. Minimum depth of coverage should be established during the test validation process and will depend upon the required sensitivity of the assay as well as the targeting/sequencing method. Regions of sequence not meeting the required read depth, especially genes with highest priority, such as pertinent genes in each tumor type, should be clearly reported as indeterminate. Of note, MAFs are not routinely reported. Medical oncologists still need to be educated for their proper interpretation, which is an important aspect of the process of implementing NGS reports in a clinical lab e continued medical education so

Table 3 e Evidence-based knowledge database with examples of predictive, prognostic or diagnostic associations (actionable) and other biologically relevant somatic cancer variants (not clearly actionable). Gene

Variant

Type

Disease

Effect

Therapeutic context

Actionable variants with predictive associations BRAF V600E Missense Melanoma

Responsive vemurafenib, dabrafenib ALK L1196M Missense Lung Resistant crizotinib ERBB2 Amplification Copy Endometrial Not trastuzumab number gain responsive single agent BRAF Y472C Missense Lung Responsive dasatinib þ erlotinib ERBB2 L755S Missense Breast Sensitive neratinib FBWX7 R367* Nonsense Breast Resistant anti-tubulin agents EGFR Exon 20 Insertion Lung Reduced erlotinib, afatinib sensitivity PDGFRA D842V Missense Gastrointestinal Reduced imatinib stromal tumor sensitivity sunitinib

Gene

Variant

Type

Variant

Type

Disease

Other variants with biological relevance RAF1 R495H Missense

Lung

FGFR2 MYC

Melanoma Colorectal

I642V Amplification

Missense Copy number gain

Level of evidence

PMID

Consensus FDA-approved FDA Emerging Emerging

Early trials Early trials

22277784 19840887

Emerging Emerging Emerging Emerging

Case report Preclinical Preclinical Preclinical

22649091 23220880 21368834 21764376

Emerging

Preclinical

15685537 16638875

Disease

Actionable variants with prognostic or diagnostic associations BRAF V600E Missense CTNNB1 T41A Missense CDH1 E243K Missense

Gene

Status

Colorectal Desmoid fibromatosis Lobular breast cancer

Association

PMID

Prognostic Diagnostic Diagnostic

23056577 23020601 23000897

Effect

Pertinence

PMID

High functional impact/ tyrosine kinase domain Loss-of-function

Pathway activation, theoretical actionability Tumor biology Tumor biology

Not available 19147536 22810696

PMID: Pubmed Identifier.


12


that physicians are trained to understand molecular profile results. Currently, MAF counts do not change clinical practice but this information is very useful in the research setting, reflecting clonal evolution and selection when NGS tests are performed in different samples and time points over the course of disease and therapy. Based on their value in specific contexts, we hope to add MAFs counts to our reports in the near future.

4.

Maintenance of a knowledge database

Once a somatic variant has been classified for clinical reporting, the curated information may be stored for future use in a knowledge database. The database should be designed to accommodate new clinical and preclinical data and may ideally be integrated with a system for automated report generation. We developed a comprehensive in-house database that contains curated annotations on the most frequent and relevant variants in multiple tumor types, according to COSMIC, TCGA and based on the publicly available resources My Cancer Genome and Targeted Cancer Care. Figure 4 summarizes some relevant findings derived from our internal knowledge database. Tumor types

with the largest number of gene variants presenting emerging predictive associations are acute leukemias, non-small cell lung cancer, brain tumors, melanoma, colorectal, ovarian and breast malignancies. A team with expertise in the NGS reporting process is responsible for routinely updating the database as new scientific and clinical knowledge becomes available, particularly in the identification of novel variants of biological importance and any variants with novel therapeutic relevance. Such information is gathered with a regular and systematic review of drug regulatory and approval status, consensus guidelines, peer-reviewed publications and clinical trial databases. Our initial experience in variant-specific annotation of genes with emerging therapeutic relevance includes 375 citations across 77 unique sources. These citations reveal scientific journals that publish associations of specific genetic variants with therapeutic relevance in relatively high volume. A routine review of these high impact journals may be considered when interpreting novel somatic variants in cancer and for database maintenance. Updated versions of our knowledge database are accessible to the public through Synapse, the collaborative cloud-based repository developed at Sage Bionetworks that allows secure data sharing and database queries, under the synapse ID ‘syn2370773’ (https://www.synapse.org/#!

Figure 4 e Statistics of internal knowledge database. Upper chart shows number of genes with emerging predictive associations in different tumor types. Citations of the same emerging clinical and preclinical associations are displayed in the lower part. Please cite this article in press as: Dienstmann, R., et al., Standardized decision support in next generation sequencing reports of somatic cancer variants, Molecular Oncology (2014), http://dx.doi.org/10.1016/j.molonc.2014.03.021


Synapse:syn2370773). Many academic groups are independently working on similar projects and we believe that an international consortium on curated cancer genomics data matching genomic aberrations to targeted therapies could have a huge clinical impact. Ideally, the information should be released as an interactive web-based tool, subjected to editing, validation and critique from the medical community.

5.

Conclusion

Next-generation sequencing assays were initially developed for research or investigational purposes but will eventually become part of cancer care. During the process of clinical implementation of NGS tests, many technical, legal and ethical challenges have to be overcome and the community is outpacing regulatory agencies’ ability to provide guidance. It is of importance to highlight that Clinical Laboratory Improvement Amendments (CLIA) or Good Clinical Laboratory Practice (GCLP) certification is required for clinical centers and consulting biotechnology companies offering NGS-based cancer diagnostic tests. Several professional societies have generated guidelines for the implementation of NGS tests, with a focus on analytical validity or patient privacy rules. Nonetheless, recommendations for use of computational tools and reporting of somatic cancer variants are still missing. In addition, laboratories need to consider a reasonable timeframe in order to provide molecular information so that it can be translated into a useful clinical tool. The biggest challenge is how to convey the amount of data obtained from NGS tests and all the information reviewed for interpretation. Effective communication of results with interactive reports can promote appropriate clinical decision-making and minimize the potential for patient harm. Unfortunately, at the present time, validated evidence on specific gene variants linked to predictive, prognostic or diagnostic associations in cancer is limited. In addition, genomics knowledge is currently ahead of our ability to therapeutically target tumors, given that many mutations identified by sequencing are either linked to unapproved drugs or are not druggable by currently available molecular therapy. Importantly, while sequencing can identify druggable targets, clinicians are often left with the task of further interpretation, treatment prioritization and decision-making in the context of additional clinical information. When the best option is to offer the patient genomic-driven clinical trials, additional logistical challenges need to be overcome, including too strict eligibility criteria in phase 1 trials or slots not available at the time of referral and geographical limitations to access drug development units. These difficulties explain why only a small number of patients are ultimately enrolled in a specific trial based on the results of NGS assays, even when actionable genomic alterations are identified in the majority of the tumor samples tested (Tran et al., 2013). Multi-institutional trial networks assessing novel agents that target specific mutations are needed in order to deal with these issues. Alternatively, when physicians and patients decide on off-label use of targeted therapies, another aspects that go beyond reimbursement concerns need to be taken into consideration. There is an inherent bias to publish positive results, making it difficult

13

to aggregate information from lots of “N-of-1” experiments. Mechanisms to annotate lack of response in this setting are missing. As proposed by Dr. Richard Schilsky at the American Society of Clinical Oncology 2013 and Worldwide Innovative Networking 2013 conferences, the ideal solution would be to create a national formulary of targeted agents against common aberrations, so that every patient receiving a matched therapy in the off-label setting can be tracked and become a “cancer information donor”. These pharmacy exchange programs could generate ever-growing databanks integrating the genomic information with therapeutic response and outcome. We envision that information derived from these registries should be added to knowledge databases such as My Cancer Genome and become readily available to oncologists worldwide, providing annotated predictive genomic markers in cancer and potentially changing the paradigm of drug approval process. In conclusion, structured reporting of clinically relevant variants may help addressing the current limitations of NGS to directly guide patient care. With standardized terminology and an expanding knowledge database, variant annotation, prioritization and clinical interpretation becomes a fluid process with potential to open new therapeutic options for cancer patients.

Conflicts of interest Authors declare no competing interests in relation to the work described.

Funding R. Dienstmann is a recipient of “La Caixa International Program for Cancer Research & Education”.

Acknowledgments Gad Getz, Ranjit Shetty, Hayley E. Robinson and Marianne Boswell for their contribution to clinical implementation of next generation sequencing at our lab. R E F E R E N C E S

Abecasis, G.R., Auton, A., Brooks, L.D., et al., 2012. An integrated map of genetic variation from 1,092 human genomes. Nature 491, 56e65. Adzhubei, I.A., Schmidt, S., Peshkin, L., et al., 2010. A method and server for predicting damaging missense mutations. Nat. Methods 7, 248e249. Ascierto, P.A., Schadendorf, D., Berking, C., et al., 2013. MEK162 for patients with advanced melanoma harbouring NRAS or Val600 BRAF mutations: a non-randomised, open-label phase 2 study. Lancet Oncol. 14, 249e256. Bose, R., Kavuri, S.M., Searleman, A.C., et al., 2013. Activating HER2 mutations in HER2 gene amplification negative breast cancer. Cancer Discov. 3, 224e237.


14


Carpten, J.D., Faber, A.L., Horn, C., et al., 2007. A transforming mutation in the pleckstrin homology domain of AKT1 in cancer. Nature 448, 439e444. Carver, B.S., Chapinski, C., Wongvipat, J., et al., 2011. Reciprocal feedback regulation of PI3K and androgen receptor signaling in PTEN-deficient prostate cancer. Cancer Cell 19, 575e586. Chen, Z., Cheng, K., Walton, Z., et al., 2012. A murine lung cancer co-clinical trial identifies genetic modifiers of therapeutic response. Nature 483, 613e617. Cooper, G.M., Shendure, J., 2011. Needles in stacks of needles: finding disease-causal variants in a wealth of genomic data. Nat. Rev. Genet. 12, 628e640. Cortes, J.E., Kantarjian, H., Shah, N.P., et al., 2012. Ponatinib in refractory Philadelphia chromosome-positive leukemias. New Engl. J. Med. 367, 2075e2088. Dahlman, K.B., Xia, J., Hutchinson, K., et al., 2012. BRAF(L597) mutations in melanoma are associated with sensitivity to MEK inhibitors. Cancer Discov. 2, 791e797. Dias-Santagata, D., Akhavanfard, S., David, S.S., et al., 2010. Rapid targeted mutational analysis of human tumours: a clinical platform to guide personalized cancer medicine. EMBO Mol. Med. 2, 146e158. Dienstmann, R., Rodon, J., Tabernero, J., 2013. Biomarker-driven patient selection for early clinical trials. Curr. Opin. Oncol. 25, 305e312. Engelman, J.A., Zejnullahu, K., Mitsudomi, T., et al., 2007. MET amplification leads to gefitinib resistance in lung cancer by activating ERBB3 signaling. Science 316, 1039e1043. Flanagan, S.E., Patch, A.M., Ellard, S., 2010. Using SIFT and PolyPhen to predict loss-of-function and gain-of-function mutations. Genet. Test Mol Biomarkers 14, 533e537. Fong, P.C., Boss, D.S., Yap, T.A., et al., 2009. Inhibition of poly(ADP-ribose) polymerase in tumors from BRCA mutation carriers. New Engl. J. Med. 361, 123e134. Frousios, K., Iliopoulos, C.S., Schlitt, T., et al., 2013. Predicting the functional consequences of non-synonymous DNA sequence variants - evaluation of bioinformatics tools and development of a consensus strategy. Genomics 102, 223e228. Fu, W., O’Connor, T.D., Jun, G., et al., 2013. Analysis of 6,515 exomes reveals the recent origin of most human proteincoding variants. Nature 493, 216e220. Garraway, L.A., Lander, E.S., 2013. Lessons from the cancer genome. Cell 153, 17e37. Gartside, M.G., Chen, H., Ibrahimi, O.A., et al., 2009. Loss-offunction fibroblast growth factor receptor-2 mutations in melanoma. Mol. Cancer Res. 7, 41e54. Gnad, F., Baucom, A., Mukhyala, K., et al., 2013. Assessment of computational methods for predicting the effects of missense mutations in human cancers. BMC Genomics 14 (Suppl. 3), S7. Gonzalez-Perez, A., Lopez-Bigas, N., 2011. Improving the assessment of the outcome of nonsynonymous SNVs with a consensus deleteriousness score, Condel. Am. J. Hum. Genet. 88, 440e449. Gozgit, J.M., Wong, M.J., Moran, L., et al., 2012. Ponatinib (AP24534), a multitargeted pan-FGFR inhibitor with activity in multiple FGFR-amplified or mutated cancer models. Mol. Cancer Ther. 11, 690e699. Green, R.C., Berg, J.S., Grody, W.W., et al., 2013. ACMG recommendations for reporting of incidental findings in clinical exome and genome sequencing. Genet. Med. 15, 565e574. Griffith, M., Griffith, O.L., Coffman, A.C., et al., 2013. DGIdb: mining the druggable genome. Nat. Methods 10, 1209e1210. Guagnano, V., Kauffmann, A., Wohrle, S., et al., 2012. FGFR genetic alterations predict for sensitivity to NVP-BGJ398, a selective pan-FGFR inhibitor. Cancer Discov. 2, 1118e1133. Heinrich, M.C., Corless, C.L., Demetri, G.D., et al., 2003. Kinase mutations and imatinib response in patients with

metastatic gastrointestinal stromal tumor. J. Clin. Oncol. 21, 4342e4349. Hofmann, I., Weiss, A., Elain, G., et al., 2012. K-RAS mutant pancreatic tumors show higher sensitivity to MEK than to PI3K inhibition in vivo. PloS One 7, e44146. Iyer, G., Hanrahan, A.J., Milowsky, M.I., et al., 2012. Genome sequencing identifies a basis for everolimus sensitivity. Science 338, 221. Jo, H., Lo, P.K., Li, Y., et al., 2011. Deactivation of Akt by a small molecule inhibitor targeting pleckstrin homology domain and facilitating Akt ubiquitination. Proc. Natl. Acad. Sci. U. S. A. 108, 6486e6491. Kumar, P., Henikoff, S., Ng, P.C., 2009. Predicting the effects of coding non-synonymous variants on protein function using the SIFT algorithm. Nat. Protoc. 4, 1073e1081. Lawrence, M.S., Stojanov, P., Mermel, C.H., et al., 2014. Discovery and saturation analysis of cancer genes across 21 tumour types. Nature 505, 495e501. Lawrence, M.S., Stojanov, P., Polak, P., et al., 2013. Mutational heterogeneity in cancer and the search for new cancerassociated genes. Nature 499, 214e218. Lubin, I.M., Caggana, M., Constantin, C., et al., 2008. Ordering molecular genetic tests and reporting results: practices in laboratory and clinical settings. J. Mol. Diagn. 10, 459e468. Mitsudomi, T., Hamajima, N., Ogawa, M., et al., 2000. Prognostic significance of p53 alterations in patients with non-small cell lung cancer: a meta-analysis. Clin. Cancer Res. 6, 4055e4063. Mogi, A., Kuwano, H., 2011. TP53 mutations in nonsmall cell lung cancer. J. Biomed. Biotechnol. 2011, 583929. Prahallad, A., Sun, C., Huang, S., et al., 2012. Unresponsiveness of colon cancer to BRAF(V600E) inhibition through feedback activation of EGFR. Nature 483, 100e103. Reva, B., Antipin, Y., Sander, C., 2011. Predicting the functional impact of protein mutations: application to cancer genomics. Nucl. Acids Res. 39, e118. Scoccianti, C., Vesin, A., Martel, G., et al., 2012. Prognostic value of TP53, KRAS and EGFR mutations in nonsmall cell lung cancer: the EUELC cohort. Eur. Resp J. 40, 177e184. Sen, B., Peng, S., Tang, X., et al., 2012. Kinase-impaired BRAF mutations in lung cancer confer sensitivity to dasatinib. Sci. Transl Med. 4, 136ra170. Sequist, L.V., Soria, J.-C., Gadgeel, S.M., et al., 2013. First-inhuman evaluation of CO-1686, an irreversible, selective, and potent tyrosine kinase inhibitor of EGFR T790M. J. Clin. Oncol. 31 abstract 2524. Serra, V., Vivancos, A., Puente, X.S., et al., 2013. Clinical response to a Lapatinib-based therapy for a Li-Fraumeni syndrome patient with a novel HER2V659E mutation. Cancer Discov. 3, 1238e1244. Shi, H., Moriceau, G., Kong, X., et al., 2012. Preexisting MEK1 exon 3 mutations in V600E/KBRAF melanomas do not confer resistance to BRAF inhibitors. Cancer Discov. 2, 414e424. Stone, A., Sutherland, R.L., Musgrove, E.A., 2012. Inhibitors of cell cycle kinases: recent advances and future prospects as cancer therapeutics. Crit. Rev. Oncog. 17, 175e198. Tran, B., Brown, A.M., Bedard, P.L., et al., 2013. Feasibility of real time next generation sequencing of cancer genes linked to drug response: results from a clinical trial. Int. J. Cancer 132, 1547e1555. Van Allen, E.M., Wagle, N., Levy, M.A., 2013. Clinical analysis and interpretation of cancer genome data. J. Clin. Oncol. 31, 1825e1833. Vogelstein, B., Papadopoulos, N., Velculescu, V.E., et al., 2013. Cancer genome landscapes. Science 339, 1546e1558. Wagle, N., Berger, M.F., Davis, M.J., et al., 2012. High-throughput detection of actionable genomic alterations in clinical tumor samples by targeted, massively parallel sequencing. Cancer Discov. 2, 82e93.



Wagle, N., Emery, C., Berger, M.F., et al., 2011. Dissecting therapeutic resistance to RAF inhibition in melanoma by tumor genomic profiling. J. Clin. Oncol. 29, 3085e3096. Watt, S., Jiao, W., Brown, A.M., et al., 2013. Clinical genomics information management software linking cancer genome sequence and clinical decisions. Genomics 102, 140e147. Whittaker, S.R., Theurillat, J.P., Van Allen, E., et al., 2013. A genome-scale RNA interference screen implicates NF1 loss in resistance to RAF inhibition. Cancer Discov. 3, 350e362. Wong, W.C., Kim, D., Carter, H., et al., 2011. CHASM and SNVBox: toolkit for detecting biologically important single nucleotide mutations in cancer. Bioinformatics 27, 2147e2148.

15

Yeh, P., Chen, H., Andrews, J., et al., 2013. DNA-Mutation Inventory to Refine and Enhance Cancer Treatment (DIRECT): a catalog of clinically relevant cancer mutations to enable genome-directed anticancer therapy. Clin. Cancer Res. 19, 1894e1901. Zhang, J., Liu, J., Sun, J., et al., 2014. Identifying driver mutations from sequencing data of heterogeneous tumors in the era of personalized genome sequencing. Brief Bioinform. 15, 244e255.


Systematic Evaluation of Sanger Validation of Next-Generation Sequencing Variants.

2 mutations in ovarian cancer - next-generation sequencing analysis of 100 cases.

Next Generation Sequencing of Pooled Samples: Guideline for Variants' Filtering.

Clinical next generation sequencing in cancer.

Next-generation sequencing identifies rare variants associated with Noonan syndrome.

Development and validation of a scalable next-generation sequencing system for assessing relevant somatic variants in solid tumors.

[Next-generation sequencing for lung cancer].

Clinical implementation of germ line cancer pharmacogenetic variants during the next-generation sequencing era.

Next generation sequencing technologies.

Participating in next generation sequencing.

Next generation sequencing technologies for next generation plant breeding.

Next-generation sequencing in African Americans with colorectal cancer.

Identification of somatic alterations in stage I lung adenocarcinomas by next-generation sequencing.

Clinical next generation sequencing of pediatric-type malignancies in adult patients identifies novel somatic aberrations.

Profiling of somatic mutations in phaeochromocytoma and paraganglioma by targeted next generation sequencing analysis.

Revealing the Complexity of Breast Cancer by Next Generation Sequencing.

Reliable Detection of Somatic Mutations in Fine Needle Aspirates of Pancreatic Cancer With Next-generation Sequencing: Implications for Surgical Management.

BRCA somatic and germline mutation detection in paraffin embedded ovarian cancers by next-generation sequencing.

Targeted next generation sequencing identifies somatic mutations and gene fusions in papillary thyroid carcinoma.

Next generation sequencing reveals the antibiotic resistant variants in the genome of Pseudomonas aeruginosa.

Pitfalls of improperly procured adjacent non-neoplastic tissue for somatic mutation analysis using next-generation sequencing.

Somatic characterization of pediatric acute myeloid leukemia using next-generation sequencing.

Insights into cancer biology through next-generation sequencing.

Somatic mosaicism of a CDKL5 mutation identified by next-generation sequencing.