Integrating Multi-Modal Cancer Data Using Deep Latent Variable Path Modelling

Cancers are commonly characterised by a complex pathology encompassing genetic, microscopic and macroscopic features, which can be probed individually using imaging and omics technologies. Integrating this data to obtain a full understanding of pathology remains challenging. We introduce a new method called Deep Latent Variable Path Modelling (DLVPM), which combines the representational power of deep learning with the capacity of path modelling to identify relationships between interacting elements in a complex system. To evaluate the capabilities of DLVPM, we initially trained a foundational model to map dependencies between SNV, Methylation, miRNA-Seq, RNA-Seq and Histological data using Breast Cancer data from The Cancer Genome Atlas (TCGA). This method exhibited superior performance in mapping associations between data types compared to classical path modelling. We additionally performed successful applications of the model to: stratify single-cell data, identify synthetic lethal interactions using CRISPR-Cas9 screens derived from cell-lines, and detect histologic-transcriptional associations using spatial transcriptomic data. Results from each of these data types can then be understood with reference to the same holistic model of illness.