> Our Work > The Work We Fund

2025 PCF Challenge Award

Predicting Off-Target Resistance to Androgen Receptor Signaling Inhibitors from Histology Slides using Foundation Models

Principal Investigators: Ekta Khurana, PhD (Weill Cornell Medicine), Iman Hajirasouliha, PhD (Weill Cornell Medicine), Yu Chen, MD, PhD (Memorial Sloan Kettering Cancer Center), Anuradha Gopalan, MD (Memorial Sloan Kettering Cancer Center)

Co-Investigators: Wassim Abida, MD, PhD (Memorial Sloan Kettering Cancer Center), Francisco Sanchez Vega, PhD (Memorial Sloan Kettering Cancer Center), Brian Robinson, MD (Weill Cornell Medicine)

Young Investigator: Elvire Roblin, PhD (Weill Cornell Medicine)

Description:

  • Untreated prostate cancers rely on the androgen receptor (AR), forming the basis for the initial efficacy of androgen deprivation therapy (ADT). Yet the disease can relapse and progress to castration-resistant prostate cancer (CRPC).
  • Alterations that reactivate AR activity represent the most common driver of CRPC, and AR signaling inhibitors (ARSIs) are used in combination with ADT. However, ARSIs can result in selective pressure, and generate AR-independent tumors. These include lineage plastic subtypes: CRPC-SCL (stem cell-like), CRPC-WNT (Wnt-dependent), and CRPC-NE (neuroendocrine).
  • The presence of these AR-independent molecular subtypes indicates an urgent need for new therapeutic strategies and methods to select patients for clinical trials for drugs targeting those subtypes. Additionally, many tumors are heterogeneous – consisting of multiple subtypes.
  • A tool that would allow the early clinical identification of prostate cancer molecular subtypes from routine pathology slides would give all patients and their clinicians the information needed to make clinical decisions toward reducing disease aggressiveness and mortality.
  • In this project, Dr. Ekta Khurana and team will use cutting edge technologies to evaluate clinical pathology slides, clinical data, and gene expression data to develop artificial intelligence (AI) algorithms that can quantitatively determine the existence of each prostate cancer molecular subtype and predict resistance to ARSIs.
  • If successful, this project will develop AI tools for predicting resistance to ARSIs that can be readily deployed in the clinic.

What this means to patients: As prostate cancer progresses and develops resistance to therapies, it becomes more variable and harder to treat. Dr. Khurana and team will develop an AI tool that can determine prostate cancer molecular subtypes using standard pathology slides and predict whether or not a patient will respond to standard of care treatments. This tool could readily be deployed in global clinics and enable improved treatment decisions and clinical trial selections to be made, ultimately improving outcomes for patients with advanced prostate cancer.