Artificial intelligence enabled Stroma-Weighted Automated Grading system to improve risk stratification in Black Men

A significant disparity in prostate cancer (PCa)-associated mortality is observed in Black men, who are twice as likely to die of prostate cancer than White men. Multiple factors contribute to Black men’s disproportionately higher cancer burden. Our group and others have demonstrated that tumor biology plays a significant role in the observed disparities, emphasizing the need for risk stratification tools for the early detection of lethal PCa phenotypes in Black men. Our long-term goal is to introduce a novel label-free Stroma Weighted Automated Gleason Grading (SWAG) system for improved risk stratification of PCa in both races and nominate molecular drives of disparity for therapeutic intervention. The overall objectives of this application are to (i) annotate H&E and Multiphoton microscopy (MPM), and second harmonic generation (SHG) images with all glandular and stromal features of PCa, developing an Artificial Intelligence (AI) enabled SWAG system for risk prediction, and (ii) to define and validate the molecular drives of racial disparity using spatial transcriptomics and multiplexed CRISPR approach in vivo. The central hypothesis is that SWAG risk prediction will accurately estimate the risk of specified prostate cancer outcomes. The rationale for this project is that the development of the SWAG system is likely to offer a strong scientific framework whereby new strategies for the early prognosis of PCa. Two specific aims will test the hypothesis: In Aim 1, we will develop a compendium of H&E and MPM/SHG images that captures all major glandular and stromal patterns of prostate cancer from 450 Black and White men’s tumors with outcome data. Next, we develop a) a machine learning (ML) tool for automated annotation of glandular and stromal features in both H&E & MPM images and develop the race-specific AI-enabled MPMSWAG system. Finally, we will test the performance of the SWAG system in risk prediction utilizing large training-validation primary cohort and secondary validation multi-institute cohorts. In Aim 2, we will perform an in-depth transcriptomic analysis of MPM-AI-identified driver (stroma-activated) regions using a spatial transcriptomics/proteomics approach and evaluate how the identified novel gene signatures contribute to adverse pathology using a multiplexed lentiviral CRISPR approach in GEM model. Our unique disparity PCa cohort and bulk gene expression data will validate the novel gene signature. The proposed research is innovative because of its focus on an AI-enabled, stroma-inclusive risk predictor tool that utilizes a label-free microscopy imaging approach. The proposed study is significant as it provides a strong scientific justification for the continued development of an MPM-SWAG inclusive, integrated model in a future iteration of these studies. Comprehensive molecular profiling and validation will provide significant functional and mechanistic insight into cancer vulnerability in Black men compared to their White counterparts. Ultimately, the combined knowledge obtained has the potential to develop innovative risk predictor tools for PCa in Black men, fostering faster adoption in routine clinical decision-making.