PathME

PathME represents a comprehensive effort to integrate and interpret multi-omics data by developing two distinct yet interconnected software tools to enhance our understanding of complex diseases like cancer at a molecular system level.

The first component, a multi-modal sparse denoising autoencoder framework coupled with sparse non-negative matrix factorization, is designed to cluster patients based on multi-omics data robustly. By leveraging pathway information, this model effectively reduces the dimensionality of omics data into a pathway and patient-specific score profile, facilitating the identification of disease subtypes characterized by distinct molecular features across various cancer datasets. This approach addresses the challenges posed by the high-dimensional nature of omics data and allows for the dissection of the specific impact of each omics feature on pathway scores. The resulting patient-specific pathway score profiles enable the robust identification of disease subgroups, providing a competitive clustering performance and offering biologically plausible insights supported by post hoc analysis.

The second component, PathMe, is a Python package that unifies pathway knowledge from three major databases into a single schema using the Biological Expression Language. This tool addresses the challenges of interoperability between pathway databases, facilitating the harmonization of these resources. PathMe enhances the modeling and comprehension of biological systems by enabling users to explore pathway crosstalk, compare consensus areas, and identify discrepancies across databases.