Ashley Westerfield, a senior bioengineering major at Stanford University, built machine learning models to predict binding affinity data for potential K-Ras inhibitors.
K-Ras is an oncogenic GTPase that has been implicated in a variety of human colonic and pancreatic cancers.
This summer at the Broad taught me how important it is to have the courage to tackle problems that everyone else considers impossible to solve. Everyone I’ve met here has a seemingly boundless amount of this scientific courage, and engaging with them this summer shaped my decision to pursue a career in research. At the Broad, no goal is too lofty, no idea too unusual, no technology too out-of-reach – and I’ve learned that this mindset is a powerful one to bring to science, health, and medicine.Endogenously responsible for controlling cell proliferation, differentiation, and development through binding to GDP/GTP, the K-Ras protein acts as a conformational switch that regulates many downstream pathways in the cell. Mutations in K-Ras – particularly the exchange of glycine for aspartic acid at the 12th amino acid (G12D) – cause the protein to become constitutively active, inducing uncontrolled cell growth. As a result, a small molecule that binds to and inactivates K-Ras can potentially halt cell growth, allowing the cancer to become more treatable by conventional cancer therapeutics like chemotherapy. Our goal is to develop K-Ras inhibitor compounds with nanomolar level affinity. To help develop such a compound, we built 2D and 3D QSAR models, a machine learning technique that allows us to predict the binding affinity (and other properties) of new compounds by analyzing the 2D and 3D structures from a library of known compounds. To validate our models, we applied the leave-one-out cross validation method, and monitored several metrics, including the r²,q², and RMSE of the known activity against the predicted activity. These models will allow us to predict the binding activity of new K-Ras inhibitors before they are synthesized, which will help us triage compounds that will bind to K-Ras with nanomolar affinity.
Project: Developing high-affinity small molecule inhibitors of mutant K-Ras using 2D and 3D QSAR models
Mentor: Alisha Caliman, Center for the Development of Therapeutics (CDoT)