Exploring interpretable machine learning
A sign at Cold Spring Harbor Laboratory. Credit: Barry Sloan
This guest essay reflects the views of Nirali Somia, a graduate student at Cold Spring Harbor Laboratory. It is part of a series of essays from current researchers at the Cold Spring Harbor Laboratory who are deeply worried that that cuts to federal funding for medical research will slow progress on finding cures for diseases and discourage young scientists from pursuing their careers in science.
I am a computational biologist interested in interpretable machine learning for genomics and health care. Interpretable artificial intelligence models are important because they reveal not just what patterns exist in data, but why those patterns matter. Think of it like this: Instead of just telling a doctor “this patient has cancer,” interpretable AI models can explain that “this patient has cancer because of these specific genetic markers we found.”
Essentially, interpretability techniques can transform mysterious computer predictions — what we call “black-box” predictions because you can’t see how the computer reached its conclusion — into more understandable insights. This transparency is especially crucial for health care, where doctors need to understand how AI arrives at diagnoses or treatment recommendations before trusting those decisions with patients’ lives. In graduate school, I want to develop AI systems that can identify disease mechanisms and new drug targets, while ensuring these models remain transparent and trustworthy for clinical use.
At Cold Spring Harbor Laboratory, I am part of PREP, the Postbaccalaureate Research Education Program. This program has been transformative in introducing me to the intersection of machine learning and biology. PREP has helped me grow skills that will be fundamental as I begin the journey of my scientific career.
Now, PREP programs no longer exist because of funding cuts to the National Institutes of Health. I wouldn’t be the scientist I am today without Cold Spring Harbor Laboratory’s PREP program. As a child of immigrants and a queer woman of color, I rarely see computational scientists who share my background or experiences. The PREP program not only gave me the foundations essential for my upcoming PhD at Harvard University and the Massachusetts Institute of Technology, but also provided a supportive community where I could envision myself thriving as a scientist.
Without these programs, I think we not only lose a transformative training program for young scientists, but we also risk losing the voices and perspectives that drive innovation and ensure science serves everyone.
Nirali Somia is a graduate student at Cold Spring Harbor Laboratory.
