Getting to the Right Patient, with the Right Drug(s) at the Right Time(s) Kenna R. Mills Shaw, Ph.D. Executive Director Khalifa Institute for Personalized Cancer Therapy

Precision Medicine and the Paradigm Shift Treatment landscape in prostate cancer. Courtesy of William K. Oh, MD. DEATH Population Based  Individual “Genotype-Phenotype” Based

Why Multiplex Testing Changes Our Approach AKT1 29 patients AKT1 1.4% 2% in breast cancer Test 2500 patients to enroll 25 patients (50%) $800 per test: 2 million dollars

Evolution of Cancer Genomic Sequencing Panels 2010-2011 2011-2012 2012 2013 2014 Within each gene, hundreds of variants may be reported within different cancers. Lack of community consensus re: number of targets to sequence; push towards exome

Rapid, Efficient Identification for Basket Trials VEGF M/T TRK M/T FGFR M/T/A RET M/T,A BRAFV00E EGFR M/A cKIT M/T FLT3 M/T?A HER2 M HER3 M CSFR1 M/T PDGFR M/T Dovitinib Neratinib Dabrafenib/Trametinib Vemurafenib Multiple histologies as single cohort Multiple histologies to be analyzed as separate cohorts for very rare aberrations M Mutation T Translocation A Amplification L Loss

Thousands of Patients Tested, How Many Get the RIGHT Drug? Cornell, Vanderbilt, Michigan, MD Anderson, International cohorts Testing between 50-20,000 genes Average 5-11% of patients with mutation in an actionable gene go on subsequent genotype matched therapy. Are we just really, really, really BAD at this?

“Right” Defined as “Matched” Genes whose products are potentially targetable with established or investigational therapeutics directly indirectly (eg inhibiting downstream signaling) 33 of 50 genes on the panel were considered “potentially actionable”

Clinical Trials Categorized by Type of “Match” Genotype-selected trials: Trials requiring a specific genomic status for enrollment eg trials for patients with BRAF V600E-mutant tumors Genotype-relevant trials: Trial with a drug that directly or indirectly targets product of an gene that is altered, but does not require biomarker for eligibility Indirectly: eg inhibiting downstream signaling Eg enrolling a patient with BRAF V600E-mutant tumors on a MEK inhibitor trial Genotype-matched trials: Genotype selected or relevant

Percentage of Patients with Mutation in ‘Actionable’ Gene 39.45% 29.60% 10.25% 20.70% Potentially actionable somatic mutations 789 39.45% Non-actionable somatic mutations 414 20.70% Likely germline variants 205 10.25% No mutation/variant 592 29.60% Total 2000 100.00%

Enrollment on Genotype Matched Trials Mutation in Potentially Actionable Gene Underwent Genomic Testing N = 2000 Genotype-matched trial after genomic testing? No (1211) Genotype-Selected Trial N = 54 Genotype-Relevant Trial N = 29 Yes (789) No (706) Yes (83) 11% of pts with mutations in actionable genes went on genotype-matched trials

Lack of Enrollment NOT Due to Lack of Available Matched Trials Increasing number of histology-agnostic or multiple histology “basket” trials Increase % patients with actionable alterations with larger platforms and copy number data

Mutations in Patients Enrolled in Genomically Matched Trials

Analysis Suggests Opportunities to Improve Patient Recruitment on Targeted Trials Insurance Denial (1) Did Patient Return to Clinic after Testing? (n=429) No (75) No (124) Poor PS (4) Treatment Elsewhere (2) Non-Investigational Treatment (17) Not Eligible (11) No Trials or Slots (4) Non-Genotype Matched Trial (5) Genotype Relevant drug Off-Protocol (1) New regimen at MDACC? Yes (354) Yes (230) Was the result mentioned in the patient’s chart? Enrolled? Yes (61) No (45) No (72) Yes (158) Yes (106) No (52) Were genotype-matched trials discussed? Could a easy-to-access decision support infrastructure, coupled with a proactive alert system and accurate clinical trial tracking improve our ability to optimize patient recruitment onto genotype-selected and genotype-driven trials? 69% had test results mentioned 67% had genotype-matched trials discussed 58% enrolled on genotype-matched trials

~13,000 Patients; ~6,000 Variants Basic annotations communicated without assessment of ‘actionability’ Actionable = Variant is likely driving tumorigenesis and there is an available therapeutic option to inhibit it. Requires knowledge of the function of the gene (ex: BRAF) Requires searching and reading through published articles within PubMed and conference abstracts for any known effects of the variant of interest (ex: BRAF V600E). Requires scouring drug databases and the published literature for drugs that may target the variant

Not All Mutations Are Equally ‘Actionable’ Phase II Imatinib in KIT-Mutant/Amplified melanoma Common in mucosal and acral melanoma Selection critical Recurrent Mutations (K642E, L576P) 40% RR Non Recurrent Mutations 0% CSD indicates melanoma arising from chronically sun-damaged skin; NA, not applicable. The melanoma subtype, KIT mutational and amplification status, the ratio of mutant to wild-type alleles as determined by their respective electropherogram peak heights, and the RECIST (Response Evaluation Criteria in Solid Tumors) response achieved for each patient treated with imatinib mesylate is shown. Amplification was defined as a KIT -to-centromere ratio of 2.5 or more, with the pattern of amplification described as uniform (homogeneous) or mixed (heterogeneous) across the tumor specimen. The best response as determined by RECIST is identified by the color of the bar. The length of the bar represents the duration of time in weeks that each patient remained receiving therapy. aMutations previously identified in melanoma8,9 and gastrointestinal stroma tumors.35,36 bTumors with mutant to wild-type allelic ratios of more than 1:1, which are those hypothesized most likely to respond to therapy with KIT inhibition. cPatients receiving treatment at the time of this report. Carvajal, R. D. et al. JAMA 2011;305:2327-2334

From Lists to Decision Support Alterations Retrieve literature, aggregate data (including unpublished internal data) Manually annotate the functional significance of the alterations Drugs Annotate drug targets, development status (FDA approved or clinical development) and therapeutic implications (if any) Clinical Trials Manually annotate the trials’ genotype/biomarker-specific selection criteria Annotate for other inclusion criteria, including tumor type, bring in NLP

Personalizedcancertherapy.org Website: 26 genes >1,408 functionally significant SNVs >13.663 SNVs reviewed Single Patient Annotation Effort:       3200 Single patients annotated Drug Curation Effort:                              1504 drugs annotated Clinical Trial Curation Effort:                3240 cancer clinical trials annotated

Functional Genomics of VUS Identifies Actionable Alterations

Decision Support in Real Time Mutation annotation inquiries and clinical trial searching can now be done in real time Dear Dr. J. Doe, Thank you for submitting your annotation request for Gene_Mutation: CDKN2A H83Q Tumor Type: Endometrial adenocarcinoma Trial Consideration (if requested): 2014-0689 Gene Alteration Allelic frequency Copy Number Variation (log2) Functional Significance Annotation Actionable gene Actionable variant   CDKN2A H83Q Not applicable Unknown This variant is of unknown functional significance. While this variant (c.249C>G p.H83Q) has not been reported in dbSNP, a synonymous alteration (c.249C>A p.H83Q) has been reported in dbSNP (rs34968276) and COSMIC. There is limited literature discussing the functional consequence of this alteration. One study predicted this alteration to be functionally significant via polyphen computational analysis (Rajasekaran et al., 2008; PMID: 18573309). However, no other literature has been found that functionally supports this prediction. This alteration is within the third ankyrin repeat of CDKN2A. ANK repeat domains are necessary for protein-protein interaction. Another codon 83 mutation (H83Y) decreases CDK4 binding, decreases protein half-life, and is defective in cell cycle arrest (Castellano et. al., 1997; PMID: 9354451; Shapiro et. al., 1995; PMID: 8521414; Yarbrough et. al., 1999; PMID: 10491434). YES Potentially Frequency Alteration cBIO COSMIC CMS50 T200 Gene_Alteration 1 / 7437 (<1%) 1 / 27234 (<1%) 0 / 3212 (0%) 0 / 1774 (0%)

34 inferred inactivating 291 variant of unknown significance Decision Support in Real Time Improves ‘Matching’ to ‘Right’ Drug All patients with identified somatic variants 188 patients with mutation(s) only in non-actionable gene 196 patients with variant in actionable gene 68 known activating 11 enrolled 4 inferred activating 7 inactivating 7 enrolled 34 inferred inactivating 1 benign 291 variant of unknown significance 22 enrolled Approximately 25% of patients with mutations in actionable genes were enrolled on clinical trials using matched therapies (~12% can be potentially enrolled - still awaiting progression)

Communicating Annotations in Real Time in Effective Reporting Format

Responses vary significantly Designing the Right Trials (to collect the Right Data) for Patients at the Right Time Responses vary significantly RECIST 30% Decrease Sosman J et al. NEJM 2012 The cancer genome is a highly inter-connected and redundant network of aberrations… we need to treat it (and design trials and support testing strategies) as if we KNOW it.

Acknowledgements Khalifa Institute for Personalized Cancer Therapy- MD Anderson Cancer Center