Leveraging ML to Classify Pathogenicity of Genetic Variants in Cancer Genomics & Precision Medicine

Tuesday, May 6, 2025

3:00 PM-5:00 PM

BIOMED PhD Thesis Defense

Title:
Leveraging Machine Learning (ML) for Classifying the Pathogenicity of Genetic Variants in Cancer Genomics and Precision Medicine

Speaker:
Ashkan Bigdeli, PhD Candidate
School of Biomedical Engineering, Science and Health Systems
Drexel University

Advisors:
Robert Babak Faryabi, PhD
Associate Professor of Pathology and Laboratory Medicine
Assistant Professor of Cancer Biology
Perelman School of Medicine
University of Pennsylvania
 
Ahmet Sacan, PhD
Teaching Professor
School of Biomedical Engineering, Science and Health Systems
Drexel University

Details:
Accurate classification of genetic variants is crucial for advancing precision medicine, particularly in oncology, where variant interpretation can guide treatment strategies and accelerate clinical research. Existing classification methods remain largely manual, inherently variable, and difficult to scale. To this end, we introduce Azurify, a machine learning-based model developed to classify the pathogenicity of somatic variants across diverse cancer types.

Trained on a broad set of cancer sequencing data and over 15,000 clinically classified variants, Azurify demonstrates robust performance across independent datasets from multiple clinical laboratories, outperforming existing classification tools and generalizing well across platforms. Azurify is designed to be easily deployed and scaled, allowing it to integrate new data as it becomes available, making Azurify adaptable for both clinical support and research environments. We demonstrate Azurify through the accurate molecular profiling of pathogenic mutations in patients with Acute Myeloid Leukemia (AML) and Chronic Lymphocytic Leukemia (CLL).

We additionally applied Azurify to sequencing data from CAR-T–treated B-cell non-Hodgkin lymphoma (B-NHL) patients, identifying key genetic predictors of treatment response and toxicity. These findings underscore the potential of machine learning to inform immunotherapy outcomes, offering a data-driven approach to precision oncology that can support patient stratification and personalized treatment strategies.

Contact Information

Natalia Broz
njb33@drexel.edu