Abstract 1072: MOSAIC Window: Spatial intra-tumoral heterogeneity characterization through large-scale spatial and single-nuclei multi-omics profiling

Abstract INTRODUCTION- MOSAIC (Multi-Omics Spatial Atlas in Cancer) is a multicenter clinical omics study aimed at deciphering cancer biology. We systematically profile thousands of cancer samples, focusing on spatial transcriptomics and single-nuclei RNA sequencing (snRNAseq) data across multiple tumor types.Here, we present the MOSAIC Window initiative, our first open science dataset. This multi-omics dataset can be leveraged to explore tumor heterogeneity and biomarker discovery within the tumor microenvironment (TME). METHODS- The MOSAIC Window dataset includes 60 samples from patients with Muscle Invasive Bladder Cancer (n = 15), Ovarian cancer (n = 15), Diffuse-large B-cell Lymphoma, Mesothelioma and Glioblastoma (n = 10 each).Each sample underwent Visium spatial transcriptomics, Chromium Flex snRNAseq, bulk RNAseq, whole-exome sequencing, and hematoxylin-eosin histology scans. We collected extensive clinical information, including age; gender; BMI; risk factors; cancer treatments; clinical, pathological and molecular information; and progression-free and overall survival.We applied strict quality control metrics in each data modality which we are reporting. Tumor heterogeneity was measured using snRNAseq and spatial transcriptomics integration. First, scANVI method was used on snRNAseq data to identify and annotate sub-populations cells from the TME. For cancer cell population, shared nearest neighbour (SNN) modularity optimization based clustering algorithm was used to separate clusters of cancer cell sub-populations. Differential expression analysis was used between these sub-populations to generate references. Then, cell type deconvolution was performed using cell2location and the references generated from snRNAseq data. RESULTS- The analysis of the spatial distribution of sub-population of cancer cells revealed distinct sub-populations of cancer cells, corresponding to different molecular sub-clones and/or TME cell composition (niches). Our computational approach highlighted significant variability in intra-tumoral heterogeneity across cancer types and patients. Samples ranged from having a single homogeneous niche to multiple distinct niches, each characterized by specific malignant cell subtypes and their associated TME, highlighting the interplay between these two compartments, which we are detailing and quantifying. These results provide important insights into potential mechanisms of treatment resistance, biomarker discovery, patient subgroup identification, and potential combination treatment strategies. CONCLUSION- The open science MOSAIC Window dataset demonstrates how integrating spatial and single-nuclei data provides key insights understanding of intra- and inter-patient tumor heterogeneity, paving the way for novel therapeutic strategies. Data is accessible on demand on EGA. Citation Format: The MOSAIC Consortium (Owkin, FR; Charite, DE; CHUV, CH; GR, FR; UKER, DE; UPMC, USA), Presenter: Caroline Hoffmann. MOSAIC Window: Spatial intra-tumoral heterogeneity characterization through large-scale spatial and single-nuclei multi-omics profiling [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 1072.