Breast cancer (BC) is the most frequent cancer and second-leading cause of cancer deaths in women in the United States. While RAS mutations are infrequent in BC, triple-negative (TN) and HER2-positive (HER2+) BC both exhibit increased RAS activity. Here, we tested the RAS effectors RALA and RALB, which are overexpressed in BC, as tractable molecular targets in these subtypes. While analysis of the breast cancer patient sample data suggests that the RALs are associated with poor outcome in both TNBC and HER2+ BC, our in vivo and in vitro experimental findings revealed the RALs to be essential in only the TNBC cell lines. While testing the response of the BC cell lines to the RAL inhibitors RBC8 and BQU57, we observed no correlation between drug efficacy and cell line dependency on RAL expression for survival, suggesting that these compounds kill via off-target effects. Finally, we report the discovery of a new small molecule inhibitor, OSURALi, which exhibits strong RAL binding, effectively inhibits RAL activation, and is significantly more toxic to RAL-dependent TNBC cells than RAL-independent HER2+ and normal cell lines. These results support the RALs as viable molecular targets in TNBC and the further investigation of OSURALi as a therapeutic agent.

DNA-encoded libraries (DELs) provide the means to make and screen millions of diverse compounds against a target of interest in a single experiment. However, despite producing large volumes of binding data at a relatively low cost, the DEL selection process is susceptible to noise, necessitating computational follow-up to increase signal-to-noise ratios. In this work, we present a set of informatics tools to analyze DEL selection data so that subsequent DEL screens probe productive regions of chemical space. Our approach segments DEL data at the individual building block level to identify productive building blocks in a library. We show how similar building blocks have a similar probability of binding, which we then employ to predict the behavior of untested building blocks. Lastly, we build a model from the inference that the combined behavior of individual building blocks is predictive of the activity of an overall compound. We report a performance of more than an order of magnitude greater than random guessing on a holdout set, demonstrating that our model can serve as a baseline for comparison against other machine learning models on DEL data.