Sindimwo Thierry, a junior molecular biology major at Pomona College, developed mutation-based learning models to predict genetic dependency for TP53 and KRAS.
Mutations, changes in DNA sequences, can lead to cancer cell proliferation. Through CRISPR knock-out screens, Project Achilles produced genetic dependency scores which reflect how genes impact cell survival. During BSRP, I was challenged to exit my comfort zone as this was my first time engaging with computational biology research. However, I believe I emerged from the program as a more confident and well-rounded scientist because of the enriching and collaborative environment that the Broad fosters. I am grateful to have spent a summer with Broadies who were so dynamic, brilliant, and inspirational! However, loss-of-function screenings are not feasible in clinical settings, highlighting the need to develop models that can effectively predict genetic dependency. Using data from Project Achilles and the Cancer Cell Line Encyclopedia, we developed mutation-based models that predict dependency scores for the genes TP53 and KRAS. We found that mutation status is an effective predictor for TP53 and KRAS dependency, although model performance was significantly more effective for KRAS. Further research should investigate the underlying biological explanations for why KRAS performs better than TP53 in mutation-based predictive models. Future efforts should also seek to improve the dependency models by incorporating other omics features such as expression and gene copy number. Ultimately, a model that effectively predicts genetic dependency scores can potentially improve gene-targeted treatments for cancer patients.
Project: Prediction of Cancer Genetic Dependency from Omics Features
Mentors: David Wu, Chris Lo, Cancer Data Science Group