A Network-Based Framework to Discover Treatment-Response-Predicting Biomarkers for Complex Diseases.

Area of Science:

• Biomedical Informatics
• Network Medicine
• Precision Medicine

Background:

• Precision medicine for complex autoimmune diseases is hindered by limited data and high-dimensional omics data.
• Developing effective treatments requires robust biomarkers for patient stratification.

Purpose of the Study:

• To introduce PRoBeNet (Predictive Response Biomarkers using Network medicine), a novel framework to identify predictive biomarkers for autoimmune disease therapies.
• To address data limitations in high-throughput multi-omics studies for precision medicine.

Main Methods:

• PRoBeNet integrates therapy-targeted proteins, disease signatures, and the human interactome network.
• Biomarker discovery and validation using retrospective and prospective patient data (ulcerative colitis, rheumatoid arthritis, Crohn disease).
• Machine learning models were built using PRoBeNet biomarkers, all genes, or random genes.

Main Results:

• PRoBeNet identified biomarkers predicting response to infliximab and a MAPK3/1 inhibitor.
• Models using PRoBeNet biomarkers significantly outperformed others, especially with limited data.
• Validated predictive biomarkers in ulcerative colitis, rheumatoid arthritis, and Crohn disease patient cohorts.

Conclusions:

• PRoBeNet effectively reduces feature dimensionality and builds robust machine learning models for precision medicine, even with scarce data.
• PRoBeNet facilitates the development of diagnostic assays for clinical trial stratification and improved patient outcomes in autoimmune diseases.