It wasn’t all beach, sun and palm trees for Nabhonil Kar '21 when he had the special opportunity to travel to the Big Island of Hawaii earlier this month. Kar made the trip to present research that overlapped his interest in biology and machine learning at the 25thannual Pacific Symposium on Biocomputing, held from January 3rd to 7th.
Kar presented his research, “Bayesian semi-nonnegative matrix tri-factorization to identify pathways associated with cancer phenotypes,” at the conference’s poster session after earning grants for the special trip from Princeton University’s Center for Statistics and Machine Learning (CSML), Department of Operations Research & Financial Engineering (ORFE) and the Office of Undergraduate Research. Besides the poster session, he also attended sessions talks, workshops and networked with researchers and practitioners at the event.
“It was an excellent experience,” said Kar, a student based in ORFE. “Given my quantitative background, I found myself learning a lot during many of the biologically-focused talks, and during my own poster session, I was excited to see audience members taking notes and asking how they might incorporate aspects from my paper into their own research.”
“The conference also gave me the opportunity to connect with professionals coming from a broad range of backgrounds, including a genetic counselor, a medical physicist, a practicing surgeon, a pathologist and of course many scientists,” Kar continued. “In a stereotypical manner, I left with more questions than answers but undoubtedly with a richer understanding of science.”
Peter Ramadge, the CSML director, said these kinds of opportunities and connections are important parts of a scholar’s development.
“Nabhonil came to us with a request for help to travel to the symposium after his paper was accepted. We thought it was a great opportunity for Nabhonil to meet researchers and practitioners, and see the different collaborative research happening in machine learning. At CSML, we are also interested in working with students to gain further experience in presenting research. This symposium provided Nabhonil that opportunity.”
Kar’s research that he presented at the symposium is based on work he performed at the Cleveland Clinic, an academic medical center based in Cleveland, Ohio. Kar, part of a research team of four, worked in the center’s department that concentrates on using machine learning tools to study clinical problems, specifically cancer. Kar performed this work during a gap year he took between his sophomore and junior year.
In the project, Kar said they developed a novel machine learning algorithm to identify biological pathways associated with certain cancer phenotypes. Pathways, Kar said, can involve how proteins interact with each other. Accurately identifying these pathways can lead to better, more reliable prognosis and treatment.
The team used The Cancer Genome Atlas (TCGA) gastric cancer and metastatic gastric cancer immunotherapy clinical-trial datasets. The results from their project showed that their method can identify important pathways associated with different cancer subtypes and immunotherapy responses, and compared to current methods, doesn’t break the inherent structure of the data and is more robust to noise. Doctors and clinicians can then use these pathways to serve as prognostic biomarkers.
In presenting his research at the symposium, Kar said it was at times a challenging but also enriching experience because many of the people he talked to were in biology, not necessarily machine learning and statistics.
“It was a constructive, interactive process talking to people,” said Kar, who found that his poster session presentations skills also improved.
For the rest of his time at Princeton, Kar said he plans on continuing his studies in machine learning with an interest in applications to finance and biology. He is currently enrolled in the Undergraduate Certificate Program in Statistics and Machine Learning, administered by CSML, as well as the certificate in Applications in Computing. He and a group of students have also started rebuilding Princeton Data Science, a club for students interested in delving into data science.