GediNET for discovering gene associations across diseases using knowledge based machine learning approach

Area of Science:

• Computational biology
• Bioinformatics
• Machine learning in medicine

Background:

• Gene discovery for disease biomarkers commonly uses machine learning and feature selection.
• Integrating prior biological knowledge enhances biomarker discovery for translational applications.

Purpose of the Study:

• To develop GediNET, a novel approach integrating prior biological knowledge into gene groups for disease association discovery.
• To enable the discovery of significant associations between diseases using gene signatures.

Main Methods:

• GediNET integrates prior biological knowledge with gene groups associated with specific diseases.
• A Grouping, Scoring, and Modelling (G-S-M) process identifies top-performing gene groups.
• Machine learning models are trained on ranked gene groups to identify disease-disease associations (DDA).

Main Results:

• GediNET identifies significant associations between diseases based on shared gene signatures.
• The approach facilitates the discovery of novel relationships between diseases.
• This facilitates improved diagnostic, prognostic, and therapeutic strategies.

Conclusions:

• GediNET offers a novel method for discovering disease-associated genes and biomarkers.
• The approach leverages prior biological knowledge and machine learning for enhanced discovery.
• GediNET can uncover novel disease-disease associations, advancing precision medicine.