Artificial Intelligence Strategies to Study Disease and Drug Response

What biomedical issue did you address in your research, and what did your studies find?

My Ph.D. research focused on developing new artificial intelligence (AI) and machine learning methodologies to analyze vast datasets of genes and their biological products alongside relevant clinical information. My goal was to use and create computational approaches that can handle complex information — for instance, datasets of genomics, transcriptomics, proteomics, and metabolomics, referred to collectively as "omics" — and to identify mutations that cause disease or can serve as prognostic biomarkers.

My projects largely utilized supervised and unsupervised machine learning techniques, as well as network analysis, to study several conditions, including major depressive disorder and suicidality. Our study, published Nature Translational Psychiatry, provided new insight about the effects of certain medications on patients with major depressive disorder who have specific genetic markers. I also looked at liver diseases known as cholangiopathies, using large omics datasets and studying the exposome, chemical exposures that may contribute to or protect from disease. From these studies, we identified a number of promising biomarkers for follow-up validation studies and enabled hypothesis generation for describing underlying disease mechanisms.

I also was interested in how AI and machine learning can be used in the clinic to help physicians. One such opportunity is with AI-driven alerts in the electronic health record that can help physicians know whether certain medications are suitable for a patient's genetic makeup. We conducted a usability study on AI-driven pharmacogenomic alerts for a specific antidepressant. We found that clinicians' preferences for alerts differed by medical specialty and experience, and that alerts with excessive details may increase clinician stress and cause delays in care. The findings may help inform approaches in designing pharmacogenomic alerts. Overall, the various studies in my Ph.D. addressed computational bottlenecks in research that uses high-throughput biological datasets and explored the new translational considerations that come with using AI in healthcare.

What aspects of Mayo's culture helped you grow as a scientist and as a thinker?

The strong integration between research and clinical care at Mayo provided an incredible opportunity to work closely with clinicians and refine the research questions according to what would be most beneficial for the patients. I was able to shadow the clinical practice in psychiatry, which provided invaluable insights into patients' lives. The graduate school's encouragement to seek external internships helped me gain experience in an industry setting, complementing the academic experience; I enjoyed completing an internship at Pfizer.

I really enjoyed teaching and mentoring students. I served as a teaching assistant for two graduate courses: Introduction to Molecular Pharmacology and AI and Data Science for Pharmacology. I also was a mentor for the Summer Undergraduate Research Fellowship (SURF) program. Both were great ways to gain mentorship and leadership skills and learn from Mayo's talented and curious community of students. Finally, Dr. Athreya's lab conducted daily lab meetings, called "coffee-walks," to practice discussing research with a broader audience in a casual setting — a critical skill moving forward.

What's next?

I will be starting as a clinical pharmacologist at AbbVie in North Chicago. My technical training in pharmacology — as well as the training to think scientifically, present research, and work closely with individuals with various approaches and expertise — has prepared me well for my next role.