Rational design of anticancer drug combinations with dynamic multidimensional input

Tumors across different patients can be understood as independent evolutionary processes of clonal Darwinian evolution under distinct therapeutic evolutionary pressures. Different therapeutic strategies disrupt evolution in distinct ways allowing the inference of the order and co-mutation patterns specifically associated to these therapies. Inferring evolutionary patterns from large cross-sectional and longitudinal therapy specific cohorts will identify specific mechanisms of drug resistance, the genetic background of these mechanisms and will inform the dynamic model of the main routes of drug evasion. First, using CAT(0) phylogenetic spaces, we will learn the statistics of phylogenetic processes associated specific drug mechanisms in breast cancer and melanoma. We conjecture that undisrupted evolutionary processes follow linear patterns and that specific therapies generate distinct branching patterns associated to number of alterations needed for relapse and effective size of the resistant population. Second, the highly branched processes associated to therapy allow to reconstruct the genetic alterations of ancestral clones allowing to order the genetic alterations. Combining cross-sectional information, one can elucidate the main routes of drug resistance, what alterations are selected under specific therapy and which is the mutational background in which they arise. As genomic data from clinical studies will be arriving we will generate first evolutionary models and integrate the results with the networks from dynamic modeling. By combining genomic data of longitudinal studies with state of the art network inference, we aim to uncover the main mechanisms of drug resistance and design combinatorial approaches.