Predicting immunological responses from clinical and molecular data to better inform cancer immunotherapy

Abstract

Cancer immunotherapies have shown remarkable success in treating previously fatal forms of cancer, but their response rates remain low, and they are associated with potentially life-threatening side effects. We used machine-learning approaches to explore two aspects of cancer immunotherapy. First, we developed methods for determining the strength of a patient's immune system prior to therapy. Examining flu vaccination responses, we identified genes and biological features that predict a healthy individual's ability to mount a response to an immunological challenge. Second, we examined the sensitivity of cancer cell lines to natural killer cells, as an initial step toward a new type of immunotherapy. Here, we predicted the sensitivity of several hundred cancer cell lines using a range of molecular data. These preliminary studies form a basis for ongoing research into the molecular and cellular aspects of immune therapeutics.