Sharing information to reconstruct patient-specific pathways in heterogeneous diseases

Area of Science:

• Computational biology
• Systems biology
• Genomics and Bioinformatics

Background:

• Abundant omics data (genomic, transcriptomic, epigenomic, proteomic) offer insights into disease drivers.
• Mapping data to protein-protein interaction networks helps identify perturbed signaling pathways.
• Integrating heterogeneous data across biological samples remains a significant challenge, especially in diseases like cancer with long-tailed mutation distributions.

Purpose of the Study:

• To develop a computational approach for inferring signaling pathways from heterogeneous biological data.
• To address the challenge of data integration across diverse patient samples.
• To improve the accuracy and robustness of pathway inference in complex diseases.

Main Methods:

• Developed a novel computational approach based on the prize-collecting Steiner forest problem.
• Utilized a network optimization algorithm to extract pathways from protein-protein interaction networks.
• Implemented a method for sharing information across samples to account for data heterogeneity.

Main Results:

• Successfully recovered signaling pathways that are conserved across samples while retaining sample-specific characteristics.
• Demonstrated improved ability to identify disrupted pathways by leveraging data from related tumors.
• Revealed patient-specific pathway perturbations in breast cancer data.

Conclusions:

• The developed computational approach effectively handles heterogeneous data for signaling pathway inference.
• Leveraging related tumor data enhances the recovery of disrupted pathways and uncovers patient-specific alterations.
• This method holds promise for understanding disease mechanisms in heterogeneous conditions like cancer.