Identification of infectious disease-associated host genes using machine learning techniques.

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Infectious Disease Research

Background:

• Multidrug resistance in microbes poses a significant global health threat.
• Identifying host genes involved in infectious diseases is crucial for understanding pathogenesis and discovering therapeutic targets.

Purpose of the Study:

• To develop and validate a computational method for identifying infectious disease-associated host genes.
• To leverage machine learning and network analysis for large-scale gene prediction.

Main Methods:

• Integrated sequence and protein-protein interaction network features.
• Employed machine learning techniques, including Deep Neural Networks (DNN), Support Vector Machine (SVM), and Random Forest (RF).
• Utilized pseudo-amino acid composition (PAAC) and network properties for feature selection.

Main Results:

• The DNN model achieved 86.33% accuracy, with 85.61% sensitivity and 86.57% specificity.
• The model demonstrated robust performance on blind (83.33% accuracy) and independent datasets.
• Identified potential host genes, with 76% of top predictions validated against experimentally verified human-pathogen protein-protein interactions (PPIs).
• Enrichment analysis revealed overlaps between infectious diseases and cancer, metabolic, and immune-related diseases, suggesting shared pathways.

Conclusions:

• This is the first computational method for identifying infectious disease-associated host genes, enabling large-scale predictions.
• While DNN is effective, simpler methods like SVM and RF perform comparably on smaller datasets.
• Findings suggest shared cellular signaling pathways between infectious, cancer, and metabolic diseases, offering potential for cross-disease therapeutic strategies.
• Identification of novel candidate genes advances understanding of disease pathogenesis and therapeutic development.