Back

MOSAIC: Intra-tumoral heterogeneity characterization through large-scale spatial and cell-resolved multi-omics profiling.

MOSAIC Consortium, ; Hoffmann, C.

2025-05-20 molecular biology
10.1101/2025.05.15.654189 bioRxiv
Show abstract

Precision oncology remains challenging due to gaps in understanding tumor biology and immunology, as well as the scarcity of consistent data across large cohorts. Here, we introduce MOSAIC (Multi-Omics Spatial Atlas in Cancer), a multi-center clinical Omics study designed to systematically profile thousands of cancer samples, with a focus on spatial and single-cell data across multiple tumor types. MOSAIC exploits recent technological advances to integrate spatial and single-cell data with complementary data modalities, including hematoxylin-eosin histology scans, bulk RNA and exome sequences, to generate comprehensive representations of cancer histology, genomics, and transcriptomics. MOSAIC also collects extensive and curated clinical information to ensure that patients meet the precise inclusion criteria for each cohort. The consortium aims to integrate all data modalities using artificial intelligence and other computational approaches in order to identify clinically-relevant biomarkers and cancer subtypes. This paper outlines the core objectives of MOSAIC, its potential impact, early proofs-of-concept, design, and experimental considerations. Additionally, we introduce the MOSAIC Window initiative, featuring the first released dataset from 60 patients, offering a glimpse into the projects groundbreaking potential. Using four selected patients, we demonstrate the power of multi-omic approaches to analyze and interpret intra-tumoral variability within and across patients, providing insights on specific drug sensitivity of cancer subpopulations and revealing potential impacts on therapeutic recommendations.

Matching journals

The top 7 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.