Diego Garay
Diego, a junior studying mathematics at Amherst College, investigated the genetic interactions in human tumors through a novel probabilistic algorithm.
Cancer arises from the gradual accumulation of somatic mutations. Some mutations tend to appear together in tumors (co-occurrence), while others rarely occur in the same tumor (mutual exclusivity). These patterns can reflect important biological relationships: co-occurrence may indicate cooperation between genes, while mutual exclusivity can result from synthetic lethality—where mutations in both genes are harmful to the tumor—or functional redundancy, where two genes act in the same pathway. However, existing methods often fail to fully account for confounding factors such as gene mutation rates, tumor mutation burden (TMB), and tumor subtype. These biases can lead to misleading signals of statistical significance.
I first started at the Ó³»´«Ã½ last summer, a part of the community college program. Coming back again for this summer made me realize that there are always new challenges and people looking to support you in any lab you’re in. Even when science feels intimidating, having people around who believe in you makes all the difference. The Ó³»´«Ã½ Summer Research Program always reminds you that you belong in the community and that the learning process never stops, all the while pushing you to grow not just as a researcher, but as a person. I am so grateful to have been a part of BSRP!
To address this, we applied a probabilistic framework that estimates the expected mutation rate of each gene in each tumor sample, adjusting for both gene- and sample-specific variation. This enables more accurate estimation of expected co-occurrence or exclusivity rates and helps reduce false positives. We applied this method to thousands of tumors from The Cancer Genome Atlas (TCGA), testing each gene pair for statistically significant interactions.
To evaluate performance, we benchmarked results against established tools such as DISCOVER, CoMEt, and WeXT. We also assessed biological relevance by checking whether significantly mutually exclusive gene pairs are enriched in shared pathways using functional interaction databases. This work aims to improve the detection of true genetic interactions in cancer, reduce methodological bias, and identify novel candidates for therapeutic targeting.
Project: Investigating Genetic Interactions in Human Tumor
Mentor: Nil Aygün, Cancer Program
