(295b) In Silico Prediction of Cancer Mechanism of Action

Currently cancer is the second leading cause of death in the US and is projected to become the leading cause in the near future. Traditional chemotherapy remains a key treatment method, offering distinct advantages over other treatment options, especially in the treatment of metastasized tumors. To develop new chemotherapeutic drugs, greater knowledge concerning the mechanisms of action (MoA) of current cancer drugs is required. By understanding how current drugs inhibit cancer growth through their underlying MoA and what molecular features affect the MoA of a drug, we can develop new and better chemotherapeutic agents.

Quantitative structure-activity relationship (QSAR) prediction of MoA allows us to relate structural features to the behavior of a chemical, thus enabling greater insight into the relationship between drug activity and molecular structure. Further, various types of cancer demonstrate varying responses to different MoAs. Classifying new potential drugs by MoA can reduce experimental testing by restricting testing candidates to only those potential drug candidates predicted to have a specified level of effectiveness. Previous research has employed QSAR predictions in an effort to predict the MoA of new, untested chemicals; however, these predictions required experimental input, usually involving the growth inhibition profiles for a panel of cancer cells lines. Our efforts have centered on advancing beyond the current state of the art by providing QSAR predictions that are truly a priori in nature.  Without any experimental knowledge, we can classify the National Cancer Institute Anti-cancer Agent Mechanism Database, which consists of 122 molecules. Our model has an overall predictive accuracy of 84%, with 10% of the molecules not classified into any MoA class and 6% of the molecules classified into an incorrect class.