Kim Ha, a senior majoring in molecular biology and biochemistry at the University of Washington, developed a statistical model to predict cancer dependency scores using omics features.
Approximately ten million people die from cancer worldwide each year, and the US National Cancer Institute estimates that 39.5% of Americans will be diagnosed with cancer in their lifetime. BSRP 2021 was such a memorable summer full of amazing people and friendships. Being able to learn from top researchers in cancer genomics and collaborate with talented peers has been so inspiring. I feel like I’ve grown so much as a scientist in such a short amount of time and I’m excited to take what I learned into the next part of my academic journey.A variety of genes regulate cancer development, including a subclass that we refer to as genetic dependencies, or genes that are required for cancer cell growth and survival. These genes are of particular interest for their therapeutic potential. Genome-wide CRISPR-Cas9 knockout screens have been a powerful tool to identify cancer genetic dependencies in cancer cell line models. However, these approaches are not directly applicable to patient tumor samples. Thus, it is important to infer cancer dependency from data that is more practical to obtain from patient tumor samples, namely, “omics” data. In this project, we assessed the power of these omics features in predicting cancer genetic dependencies. To answer this question, we selected omics features such as copy number, mutational status, and gene expression from the Broad Institute’s Cancer Cell Line Encyclopedia (CCLE) dataset to build a multivariable linear regression models to predict CCLE cell line genetic dependency data from the Broad’s Project Achilles. From our analysis of the copy number feature, we found that this feature produced models with strong predictive power for two of five gene dependencies tested which are KRAS and EGFR. This suggests that gene copy number may be a helpful feature in predicting cancer dependency scores, but additional testing on more gene dependencies is needed. We anticipate that our statistical analysis will inform future research related to the development of therapeutic agents for cancer.
Project: Predicting Cancer Dependency from Omics Features
Mentors: David Wu, Cancer Data Science Team
Krinio Giannikou, Kwiatkowski Lab