MOSAIC: Multi-Omic Spatial Atlas in Cancer, effect on precision oncology.

Abstract

e15076 Background: Precision oncology aims to first break diagnoses into biologically distinct subtypes, then pursue personalized therapies for each group of patients ( Garraway et al, J Clin Onc, 2013). This strategy is enabled by recent advances in technologies for medical imaging, molecular profiling, and artificial intelligence. Spatially resolved molecular profiling (“spatial omics”) is an emerging technology that harnesses all three of these advances ( Navarro et al, Science, 2016; Rao et al, Nature, 2021). Cancer is driven by localized interactions between tumor cells, immune and non-immune stromal components. Methods: Spatial omics studies have revealed specific cell types and pathways that control those tumor-microenvironment crosstalks ( Hunter et al, Nat Comm, 2021; Alon et al, Science, 2021), involving both innate and adaptive immunity ( Binnewies et al, Nat Med, 2018). Identifying new cancer subtypes based on the spatially localized tumor-host interactions across many patients have the potential to revolutionize immuno-oncology. The advances in utilizing AI-based methods have so far been unattainable due to a small number of patients in spatial omics studies. Results: Here we present the MOSAIC project - a collaborative initiative across industry and top oncology hospitals to build the largest collection of spatial omics data in cancer. We combine comprehensive clinical annotation with new deep profiling methods to both discover cancer subtypes and identify drug targets and biomarkers within them. MOSAIC aims to generate multimodal data for a total of 7,000 patients in seven cancer indications. Data modalities will include spatial and single cell transcriptomics and proteomics, bulk molecular profiling, pathology images, and curated clinical information. This collaboration harnesses the strengths of academia and industry to provide patient samples, generate high-quality data, develop AI-based analytical tools, and compile a resource that will eventually be made widely available for medical research. Conclusions: Here we present the workflow of the MOSAIC study, demonstrate the data processing pipelines and show spatial distributions of transcripts within the tumor microenvironment. Building upon these initial datasets, we discuss how the MOSAIC parties will collaborate to build an unprecedented database for research and discovery of new immuno-oncology therapeutic approaches.