DeepPFP-CO

DeepPFP-CO predicts protein functions by modeling co-occurrence relationships among Gene Ontology (GO) terms with deep learning to improve functional annotation for applications such as drug discovery and agricultural biotechnology.

Key Features:

• Graph Convolutional Network (GCN) Integration: Employs a Graph Convolutional Network to capture and utilize co-occurrence relationships between Gene Ontology (GO) terms.
• Co-Occurrence Exploration: Integrates the predicted propensity of central GO functions with their associated co-occurring functions to produce comprehensive functional profiles for proteins.
• Performance Metrics: Evaluates efficacy using Fmax and Area Under the Precision-Recall Curve (AUPR) and compares results with methods such as DeepGOPlus and DeepGOA.
• Protein-Level Analysis: Performs analyses at the protein level in addition to individual GO term predictions to validate predictive capabilities.

Scientific Applications:

• Drug Discovery: Enables more accurate identification of potential drug targets through improved protein function annotation.
• Agricultural Biotechnology: Supports functional characterization of proteins relevant to crop trait development and improvement.

Methodology:

Uses deep learning, specifically a Graph Convolutional Network, to model co-occurrence relationships among GO terms and integrates predicted propensities of central functions with associated co-occurring functions.