PhotoModPlus

PhotoModPlus predicts photosynthetic protein functions in prokaryotic genomes using genome neighborhood networks and machine learning to support functional annotation and evolutionary analysis.

Key Features:

• Genome Neighborhood Networks (GNN): Uses GNNs to visualize conserved neighboring genes across multiple photosynthetic prokaryotic genomes to reveal genomic context for queries.
• PhotoModGO multi-label classification: Employs the PhotoModGO model that uses genome neighborhood features to predict photosynthesis-specific functions across 24 prokaryotic photosynthesis-related Gene Ontology (GO) terms.
• Validation and performance: Achieves an F1 measure up to 0.872 as assessed by nested five-fold cross-validation and outperforms traditional sequence-based approaches.
• Context-aware prediction: Focuses on genome neighborhood features rather than sequence-only data to provide context-aware functional predictions.

Scientific Applications:

• Identification of photosynthetic proteins: Integrates GNN visualization and PhotoModGO predictions to identify novel photosynthetic proteins in prokaryotic genomes.
• Functional guidance and evolutionary insights: Uses conserved genomic neighborhood analysis to provide functional annotation guidance and insights into the evolutionary dynamics of photosynthesis-related genes.

Methodology:

Applies genome neighborhood networks for visualization, extracts genome neighborhood features to input into the PhotoModGO multi-label classifier predicting 24 photosynthesis-related GO terms, and evaluates performance by nested five-fold cross-validation compared against sequence-based approaches.