iHRD: An integrated genomic framework for classifying homologous recombination DNA repair deficiency (HRD) in prostate cancers and predicting responses to therapeutics targeting HRD

No author information is available for this submission.

70 views
14 downloads

Presentation

The authors are in the process of completing this submission.

Please try again later.

Abstract

Authors:
Navonil De Sarkar, Sayan Dasgupta, Ilsa Coleman and Peter S Nelson

Affiliations:
Fred Hutchinson Cancer Research Center

Abstract Body:
"Background:
Metastatic castration-resistant prostate cancer (mCRPC) comprises a spectrum of subtypes defined by recurrent molecular aberrations. Several subtypes are notable for particular vulnerabilities that are associated with exceptional responses to specific therapies. Homologous recombination DNA repair deficiency (HRD) in a tumor sensitizes it to PARPi/Platinum therapies. Currently, the allocation of these agents is based on the identification of mutations in HR-associated genes; however, this approach for HRD prediction is only 60% efficient. Improved biomarkers of functional HRD are needed to optimize treatment. and impact mCRPC outcomes.

Methods:
We adopted an empirical approach to define HRD based on high penetrance and low penetrance genomic features and constructed an integrative classification framework named iHRD. iHRD adopts a radial basis function kernel in performing SVM binary classification. We performed empirical learning from whole exome sequences (WES) generated from 48 bonafide bi-allelic HR pathway gene mutants and trained the classifier against WES from 187 tumors without HR gene mutations. We then determined the iHRD classification for 418 SU2C tumors and an independent mCRPC cohort consisting of 139 tumors (Model parameter: Gamma=0.01, cost=1). We also determined the iHRD status for 19 prostate cancer PDX models and evaluated carboplatin sensitivity to validate the utility of iHRD classification.

Results:
The iHRD model associated with biallelic HR gene alterations could attain 98.36% efficiency. Overall, 91.6% of tumors with biallelic HR gene alterations in the SU2C case series were classified as iHRD(+). Most ATM and CHEK2 mutants were iHRD(-). All bi-allelic NHEJ pathway gene mutants, bi-allelic BER pathway gene mutants, and bi-allelic TP53BP1, REV7, and MUS81 mutants were classified as iHRD(-). All hypermutated tumors, independent of their HR pathway function status were classified as iHRD(-). We determined that ~13% of tumors without any genomic aberration in HR gene mutations were classified as iHRD(+), and these included a major subset of tumors with monoallelic alterations in multiple HR genes. We identified four iHRD(+) PDX lines without genomic HR gene alterations. These tumors demonstrated substantial sensitivity to carboplatin in vitro and in vivo, which was equivalent to PDX lines with biallelic HR gene mutations. In a retrospective analysis of 22 mCRPC patients treated with PARPi, 70% of iHRD(+) responded to treatment, whereas 50% of patients with biallelic HR gene loss responded. In another retrospective analysis of 15 mCRPC cohorts that received carboplatin, iHRD(+) tumors were significantly enriched with cases who attained PSA50 response level(p=0.008).

Conclusions:
iHRD is more accurate in identifying mCRPC tumors with functional HRD and predicting responses to therapeutics that exploit HRD."

Funding Acknowledgments:
DOD W81XWH-17-1-0380 and 2019 Roxann Taylor, Michael Deaddio, & Thomas H. Lee – PCF VAlor Young Investigator Award

Corresponding Author Email:
ndesarka@fredhutch.org

Conflict of Interest:
Nothing to Disclose