AI-Guided Single-Cell Transfer Learning Model to Predict Precision Medicine in Ovarian Cancer

Project summary

Ovarian cancer patients are often diagnosed at an advanced stage and develop resistance to standard-of-care therapy. Recent studies have produced clinical drug predictions by machine learning, however, insufficient patient numbers and lack of knowledge of drug targets limit their use. Our goal is to accelerate the discovery of innovative, actionable therapies through integrative network-based AI models in an individualized manner.

Specifically, we will first develop and optimize multiomics-based biomarkers to predict clinical outcomes to standard-of-care therapies. For most of the non-responders to conventional treatment, we will then establish and validate a novel AI framework to identify actionable drugs for personalized therapy by “transfer learning” using integrated single-cell data. By AI-guided transfer learning, we aim to contextualize ovarian cancer patients’ tumor cells to best match cell lines with known susceptibility to drugs and drug combinations.

We will further integrate clinical features with multi-omics data to predict new treatment options for ovarian cancer patients by AI. Taken together, this proposal develops systems-level network-based AI models to prioritize personalized therapies based on tumor compositions at various resolution levels. This collaborative teamwork is only made possible due to availability of comprehensive clinical data across diverse populations of ovarian cancer patients from the Ovarian Cancer Moon Shot program at MD Anderson Cancer Center.

This grant was made possible in part by a generous donation from Hope for Heather in memory of Heather Weeks.