UC San Diego

Given the amount of targeted cancer therapies available in modern oncology, it comes as a surprise that many cancer patients struggle with receiving effective treatment. We have developed a novel visual paradigm consisting of Bayes’ theorem of conditional probability alongside posterior log odds calculations in order to predict the weight of evidence of individual tumor features on the efficacy of targeted therapies. We introduce the evidence plot, a modified naive Bayesian nomogram that provides higher accessibility to patients by featuring a gauge-like design, modular feature inputs, and alternative color palette options. When tested with oncological datasets, the evidence plot attained area under the receiving operator characteristic (AUROC) scores in the range of 0.825 to 0.965 for predicting success and survival. With further testing, the evidence plot can serve as a comprehensive prognostic tool for treating cancer on an individual patient basis.