DiSMVC: a multi-view graph collaborative learning framework for measuring disease similarity.

Area of Science:

• Computational biology
• Bioinformatics
• Genomics

Background:

• Understanding disease associations is crucial for identifying biomarkers and drug targets.
• Existing computational methods for disease similarity lack biological interpretability and efficiency due to limited consideration of multi-molecule regulation.

Purpose of the Study:

• To propose DiSMVC, a novel computational method for measuring disease similarity.
• To improve the biological interpretability and efficiency of disease association pattern capture.

Main Methods:

• DiSMVC utilizes a supervised graph collaborative framework.
• It integrates gene and miRNA associations via cross-view graph contrastive learning for disease representation.
• Disease similarity is computed using association pattern joint learning with phenotype data.

Main Results:

• DiSMVC effectively extracts discriminative characteristics for disease pairs.
• The method outperforms existing state-of-the-art approaches in predicting disease associations.
• Experimental results demonstrate DiSMVC's potential for molecular interpretability.

Conclusions:

• DiSMVC offers a promising approach for predicting disease associations.
• The method provides enhanced molecular interpretability compared to previous computational tools.
• DiSMVC facilitates a deeper understanding of disease pathological mechanisms.