Using artificial intelligence to discover spatial, genomic, and pathologic biomarkers to guide and augment immune checkpoint inhibitor therapy for gastric cancer - PROJECT SUMMARY/ABSTRACT Outcomes for patients with gastric cancer, which is the fourth-most lethal cancer worldwide, are poor and improved therapies are needed. While immune checkpoint inhibitors (ICIs) extend survival for patients suffering from a variety of cancers, most gastric cancer patients do not respond to ICIs. The overall low response of gastric cancers to ICI therapy is not only disappointing in terms of inability to improve overall survival, but it also means that a large portion of patients treated with ICIs suffer the toxicities of the treatment without any clinical benefit. Thus, to improve outcomes for gastric cancer patients, there is a critical need to identify novel biomarkers that guide ICI use and to find new strategies that augment ICI efficacy. The investigators’ long- term goal is to gain a deeper understanding of the gastric cancer tumor microenvironment to refine the use of ICIs. The objective of this proposal is to identify biomarkers that predict ICI response and to discover targets that are actionable to improve ICI efficacy. Based on the investigative team’s published and unpublished results, the central hypothesis of the proposal is that artificial intelligence can analyze transcriptomic, digital pathology, and spatial data to guide ICI use for patients with gastric cancer. To test this hypothesis, a close collaboration between a computational scientist and surgeon-scientist has been established to pursue 3 aims. In Aim 1, we will identify novel transcriptomic, digital pathology, and spatial biomarkers that predict ICI response by comparing tumor samples obtained from gastric cancer patients who did and did not respond to ICIs. In Aim 2, we will use explainable machine learning approaches to analyze multimodal datasets to predict ICI response. In Aim 3, we will perform preclinical testing of therapies that target pathways and cell-cell interactions that are enriched in ICI nonresponders in an effort to increase ICI efficacy. Three candidate targets have already been identified. The conceptual innovations of this proposal are 1) that artificial intelligence can analyze multiple streams of data to discover predictive biomarkers, identify ICI response mechanisms, and predict ICI response, and 2) there are actionable mediators of ICI nonresponse that are identifiable via detailed analysis of the tumor microenvironment. The proposal is also supported by multiple technical innovations that include the use of 1) cutting-edge spatial profiling techniques to simultaneously acquire both spatial and transcriptomic information within the tumor microenvironment, 2) novel artificial intelligence algorithms that identify image-based predictive biomarkers, 3) novel artificial intelligence algorithms to integrate and process high-dimensional datasets and provide practical guidance on the probability of ICI response, and 4) a novel gastric cancer organoid platform to test candidate therapies to improve ICI response. The expected results from the integrated analyses of clinical, computational, and experimental data will be impactful as they will provide novel insights into gastric cancer biology and have the potential to improve both quality of life and survival outcomes for patients with gastric cancer.
