ngphubinh

ngphubinh predicts adaptor protein functions by integrating recurrent neural networks (RNNs) with position specific scoring matrix (PSSM) profiles to capture sequential and evolutionary information for protein function prediction.

Key Features:

• Input features: Uses position specific scoring matrix (PSSM) profiles to provide evolutionary information about protein sequences.
• Model architecture: Integrates recurrent neural networks (RNNs) to capture sequential dependencies in PSSM-derived features.
• Target biological entities: Focuses on adaptor proteins involved in signal transduction, characterized by modular domains and binding activities.
• Addressed limitation: Mitigates the loss of sequential information inherent in using PSSM features alone by applying RNNs.
• Evaluation metrics: Reports area under the receiver operating characteristic curve (AUC) of 0.893 (cross-validation) and 0.853 (independent dataset).
• Benchmarking: Demonstrates superior performance relative to existing state-of-the-art methods.

Scientific Applications:

• Adaptor protein annotation: Predicts functional classes of adaptor proteins to aid annotation in signaling pathways.
• Protein function prediction: Enhances computational prediction of protein function by combining evolutionary profiles with sequence-aware models.
• Computational biology research: Serves as a methodological basis for exploring RNNs with PSSM profiles in sequence analysis and method development.

Methodology:

Combines position specific scoring matrix (PSSM) profiles as input features with recurrent neural networks (RNNs) to capture sequential dependencies and evaluates performance using cross-validation and independent datasets reporting AUCs of 0.893 and 0.853.