My work develops statistical and machine learning methods for prediction in biomedical applications, particularly for high-dimensional and heterogeneous data. I enjoy working with a broad range of modeling approaches, from sparse linear models to large protein language models. Interpretability and ease-of-use are important to me, and I focus on designing the simplest models that achieve high performance.
My research applications are most often in cancer, immunology, and their intersection, and I work across the spectrum of biomedical data, including biological data (RNA/DNA/proteins) and electronic health records.
Please describe one or two of your most interesting projects.
Cancer patients who relapse following treatment have ‘minimal residual disease’ (MRD) cells. We know how to characterize these cells after treatment, but we do not know what they look like at diagnosis. If we could identify MRD cells at diagnosis, we could offer a stronger first-line therapy and design new drugs to target these cells. In retrospective studies with single-cell diagnostic samples from cancer patients, we typically know which patients relapsed, but not which individual cells are MRD. To tackle this, we developed Mixture Modeling for Multiple Instance Learning (MMIL), a method that trains models to identify MRD cells using only patient-level labels. We found that MMIL had strong performance predicting which patients would go on to relapse, and it offered insight into features of cancer cells which might contribute to chemoresistance. We additionally found that MMIL can be a useful tool even when gold-standard labels are available: for example, it can aid hematopathologists as they perform blast enumeration by manually counting cells under a microscope.
What makes you excited about your data science and AI research?
I love finding solutions to challenging modeling questions: doing biomedical data science is like solving puzzles with the goal to positively impact the world.
I also love collaboration. My colleagues in biomedicine (biologists, chemists, clinicians) teach me exciting science, and we get to put our heads together to make the world a happier, healthier place.