ILMF-VH

ILMF-VH predicts potential associations between viruses and their hosts to elucidate virus-host interaction networks and their implications in microbial ecology and human disease.

Key Features:

• Integration of Known Associations: Incorporates known virus-host association networks into the predictive model.
• Virus Network: Builds a virus network using oligonucleotide frequency measurements to capture viral genomic characteristics.
• Host Network (Similarity Network Fusion): Constructs a host network by integrating oligonucleotide frequency similarity and Gaussian interaction profile kernel similarity via Similarity Network Fusion.
• Kernelized Logistic Matrix Factorization (KLMF): Applies KLMF to integrate diverse biological data on the heterogeneous network for prediction of virus-host interactions.
• Enhanced Prediction Accuracy: Empirical evaluations report superior host prediction accuracy compared to other existing methods.

Scientific Applications:

• Understanding Microbial Dynamics: Predicts virus-host interactions to help elucidate functional networks within microbial communities.
• Discovery of Novel Hosts: Identifies previously unrecognized host-virus associations, exemplified by prediction of Escherichia coli as a potential host for crAssphage.

Methodology:

Constructs separate virus and host networks (virus: oligonucleotide frequency; host: oligonucleotide frequency similarity and Gaussian interaction profile kernel similarity fused via Similarity Network Fusion) and applies Kernelized Logistic Matrix Factorization on the combined heterogeneous network to predict virus-host associations.