Presentation Detail
Presenter
Ilona Holcomb, PhD, Director of Bioscience, ImpriMed, Inc.
Research Abstract
Introduction
Patients with the same cancer diagnosis may respond differently to anticancer drugs. ImpriMed helps caregivers find effective treatments by providing personalized clinical outcome predictions for commonly used drugs. Our predictions are generated by artificial intelligence (AI) models trained on clinical data and functional and molecular biomarkers. We present here a pilot study designed to increase the power of ImpriMed’s commercial AI models for canine B-cell lymphoma (BCL) using NGS technology.
Methods
From ImpriMed’s biobank of 1,000 paired tumor/normal canine lymphoma DNA samples with known clinical outcomes, we selected 48 patients for a pilot study. Mutation profiles were generated for each patient using targeted NGS of 578 genes selected based on their association with cancer.
Results
We identified 377 variants/patient on average (71 somatic, 306 germline), including known BCL-associated variants like FBXW7 (13%) and MDR1(4%). Genes were ranked by degree of correlation with clinical CR induced by CHOP therapy and by time to relapse after CHOP. We identified multiple novel candidate biomarkers, including a gene we refer to as Predictor of Slow Relapse-1 (PSR1). Variants in PSR1, a signal transduction factor and putative tumor suppressor in human BCL, are striking correlated with time to relapse in BCL patients (0.61 Pearson r).
Conclusion
ImpriMed’s new cancer gene panel and NGS pipeline, used in conjunction with our clinical outcomes database, provide a powerful platform for identifying novel predictive biomarkers of the anticancer drug response. This platform is now being used to collect additional data for more powerful commercial AI models.
