Integrative Prediction of Therapeutic Response in T-cell Lymphoma by Omic and Spatial Modeling - ABSTRACT The tumor ecosystem plays a critical role in tumor development, progression and therapeutic response. Previous studies have utilized dissociative and single-cell omics technologies to profile the tumor ecosystem, specifically to understand therapeutic resistance and identify predictive biomarkers for precision cancer medicine. Yet, very few of these biomarkers have adequate performance characteristics for adoption in clinical practice. We hypothesize that a fundamental facet of the tumor ecosystem, i.e., the spatial organization of cells, which encodes key information involving paracrine and juxtracrine interactions that drive “neighborhood- level” biology, can further inform predictive models. Recent technological breakthroughs in highly multiplexed imaging and spatial transcriptomics offer an unprecedented opportunity to delineate the therapeutic consequences of spatial relationships within clinical tumor samples. Quantitative spatial features can provide independent valuable information, which is unlikely to be captured by clinical, genetic and bulk-transcriptional predictors. Hence, we propose to integrate highly multiplexed imaging data with omic approaches to delineate mechanisms of resistance and build predictive models of response for patients with T-cell lymphoma, who have a desperate unmet clinical need. In Aim 1 (K99 phase), I will build automated computational tools to robustly quantify spatial features from highly multiplexed imaging data and integrate it with exome and RNA- Seq. I will utilize >100 primary specimens collected pre-, on- and after-treatment with the PI3K-δγ inhibitor duvelisib to nominate mechanisms of de novo and acquired resistance. In Aim 2 (K99 phase), I will build an integrated machine-learning model to predict which patients are most likely to benefit from duvelisib and evaluate the impact of spatial features towards model performance. In Aim 3 (R00 phase), I will validate the model in an independent cohort and extend to samples from patients treated with additional agents, to identify consistent and parsimonious signatures of spatial features that could be developed for broader use. My extensive background in computational biology and experimental biology puts me in a unique position to accomplish this proposal. During the K99 phase, I will be supported by an outstanding and interdisciplinary team of advisors and collaborators (Drs. David Weinstock, Peter Sorger, Jon Aster, Allon Klein, Peter Park, and Steven Horwitz) with expertise in all aspects of the proposed research. I will acquire new skills in (1) computational analysis of highly multiplexed imaging to model molecular and spatial information, (2) data integration methods to delineate regulatory programs for designing effective drug combinations and (3) analysis of predictive biomarkers in clinical trial samples from clinical trials. Together with institutional support from Dana Farber Cancer Center and formal coursework and training, I will bridge my knowledge gap in cancer biology and gain the communication and leadership skills vital to transition into an independent position and establish an independent, data science-driven, translational research program.
