A network embedding model for pathogenic genes prediction by multi-path random walking on heterogeneous network.

Area of Science:

• Computational biology
• Genetics
• Bioinformatics

Background:

• Predicting pathogenic genes is vital for disease management but traditional methods are slow and complex.
• Computational biology offers a more efficient approach to identifying candidate pathogenic genes.

Purpose of the Study:

• To introduce Multipath2vec, a network embedding method for enhanced pathogenic gene prediction.
• To improve upon existing computational methods for identifying disease-associated genes.

Main Methods:

• Constructed a heterogeneous GP-network integrating gene-phenotype relationships and interactions.
• Developed a multi-path random walk strategy to capture complex network structures.
• Utilized learned vector representations to calculate gene-phenotype similarities for ranking candidate genes.

Main Results:

• Multipath2vec demonstrated superior performance compared to baseline methods (CATAPULT, PRINCE, Deepwalk, Metapath2vec).
• The method was validated on diverse gene-phenotype datasets, including many-genes, single-gene, and whole datasets.
• Experimental results confirmed the higher accuracy of Multipath2vec in pathogenic gene prediction.

Conclusions:

• Multipath2vec offers a highly accurate computational approach for pathogenic gene prediction.
• This method advances the field of computational biology in disease gene discovery.