Integrating Prior Knowledge Into Prognostic Biomarker Discovery based on Network Structure

TL;DR

This paper introduces FrSVM, a new algorithm that integrates protein interaction networks into gene selection, improving prognostic biomarker prediction stability and biological relevance.

Contribution

The paper presents FrSVM, a simple filter-based method that enhances gene signature stability and prediction accuracy by incorporating network information and disease-related data.

Findings

FrSVM outperforms competing methods in prediction accuracy.

Gene signatures from FrSVM are enriched with known disease genes.

Incorporating network and disease data improves biomarker relevance.

Abstract

Background: Predictive, stable and interpretable gene signatures are generally seen as an important step towards a better personalized medicine. During the last decade various methods have been proposed for that purpose. However, one important obstacle for making gene signatures a standard tool in clinics is the typical low reproducibility of these signatures combined with the difficulty to achieve a clear biological interpretation. For that purpose in the last years there has been a growing interest in approaches that try to integrate information from molecular interaction networks. Results: We propose a novel algorithm, called FrSVM, which integrates protein-protein interaction network information into gene selection for prognostic biomarker discovery. Our method is a simple filter based approach, which focuses on central genes with large differences in their expression. Compared to several other competing methods our algorithm reveals a significantly better prediction performance and higher signature stability. More- over, obtained gene lists are highly enriched with known disease genes and drug targets. We extendd our approach further by integrating information on candidate disease genes and targets of disease associated Transcript Factors (TFs).