Spatial omics technologies, generating high-throughput and multimodal data, have necessitated the development of advanced data integration methods to facilitate comprehensive biological and clinical treatment discoveries. Based on the cross-attention concept, we developed an AI learning based toolchain called StereoMM, a graph based fusion model that can incorporate omics data such as gene expression, histological images, and spatial location. StereoMM uses an attention module for omics data interaction and a graph autoencoder to integrate spatial positions and omics data in a self-supervised manner. Applying StereoMM across various cancer types and platforms has demonstrated its robust capability. StereoMM outperforms competitors in identifying spatial regions reflecting tumour progression and shows promise in classifying colorectal cancer patients into deficient mismatch repair and proficient mismatch repair groups. The comprehensive inter-modal integration and efficiency of StereoMM enable researchers to construct spatial views of integrated multimodal features efficiently, advancing thorough tissue and patient characterization.

MOTIVATION: Computational reconstruction of clonal evolution in cancers has become a crucial tool for understanding how tumors initiate and progress and how this process varies across patients. The field still struggles, however, with special challenges of applying phylogenetic methods to cancers, such as the prevalence and importance of copy number alteration (CNA) and structural variation events in tumor evolution, which are difficult to profile accurately by prevailing sequencing methods in such a way that subsequent reconstruction by phylogenetic inference algorithms is accurate. RESULTS: In this work, we develop computational methods to combine sequencing with multiplex interphase fluorescence in situ hybridization to exploit the complementary advantages of each technology in inferring accurate models of clonal CNA evolution accounting for both focal changes and aneuploidy at whole-genome scales. By integrating such information in an integer linear programming framework, we demonstrate on simulated data that incorporation of FISH data substantially improves accurate inference of focal CNA and ploidy changes in clonal evolution from deconvolving bulk sequence data. Analysis of real glioblastoma data for which FISH, bulk sequence and single cell sequence are all available confirms the power of FISH to enhance accurate reconstruction of clonal copy number evolution in conjunction with bulk and optionally single-cell sequence data. AVAILABILITY AND IMPLEMENTATION: Source code is available on Github at https://github.com/CMUSchwartzLab/FISH_deconvolution. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.