DRIVE-KG: Enhancing variant-phenotype association discovery in understudied complex diseases using heterogeneous...

Area of Science:

• Genomics and Bioinformatics
• Women's Health Research
• Computational Biology

Background:

• Endometriosis is a prevalent yet understudied women's health condition affecting 10% of reproductive-age women.
• Limited genetic characterization of endometriosis, with current GWAS explaining only 11% of heritability, necessitates integrative approaches.
• Graph representations offer a powerful framework for harmonizing diverse biological data.

Purpose of the Study:

• To present DRIVE-KG, a novel knowledge graph for integrating multi-omics data to study complex diseases like endometriosis.
• To develop and evaluate machine learning models for disease risk inference and patient-level classification using DRIVE-KG.
• To uncover novel single nucleotide polymorphism (SNP)-phenotype associations and improve diagnostic capabilities for endometriosis.

Main Methods:

• Constructed a heterogeneous graph (DRIVE-KG) integrating diverse multi-omics datasets.
• Trained a link prediction model to identify SNP-phenotype associations (endometriosis, obesity).
• Developed a graph convolutional network (GCN) for patient-level classification of endometriosis/adenomyosis.

Main Results:

• Identified 66 high-confidence candidate SNP-endometriosis associations, enriched for obesity and depressive disorder traits.
• Discovered novel genetic signals for endometriosis, distinct from known obesity associations.
• Achieved an F1 score of 0.752 for endometriosis/adenomyosis classification using GCN, outperforming genetic risk scores (0.698).
• GCN demonstrated meaningful stratification of disease severity and adenomyosis signal.

Conclusions:

• Heterogeneous integration of multi-omics data via DRIVE-KG is effective for discovery and clinical prediction in understudied diseases.
• DRIVE-KG facilitates uncovering novel genetic insights into endometriosis etiology.
• The developed GCN model shows potential for improving endometriosis/adenomyosis diagnosis and stratification.