Directed Discovery of Novel Drug Cocktails
John H. Miller, Ralph Zinner, and Brittany Barrett
Key Words:
Nonlinear search, lung carcinoma, cancer, chemotherapy, drug cocktails
Combinations of drugs can result in effective treatments for certain diseases like HIV/AIDS. Unfortunately, our ability to discover such combinations is quite limited, as drugs often interact in highly nonlinear ways and thus it is difficult to predict a priori which cocktails are likely to be effective. Moreover, the brute-force approach of screening all possible combinations fails due to the combinatorial explosion of possible cocktails, even when we consider modest numbers of candidate drugs. As an alternative, here we use a nonlinear search algorithm designed to direct the discovery of novel, effective drug cocktails. We demonstrate this approach by finding chemotherapy cocktails that can inhibit A549 (lung carcinoma) cells using a hybrid, nonlinear-optimization algorithm. We find that directed discovery can be an effective means of automatically deriving novel cocktails using a relatively small number of experiments. The basic idea of directed discovery explored here has a variety of other applications across many fields.