IRC-Fuse

IRC-Fuse predicts redox-sensitive cysteine (RSC) residues in proteins by integrating fused multiple feature representations and machine learning to produce sequence-based probability scores for RSC identification.

Key Features:

• Fused multiple feature representations: Integrates multiple feature representations into a fused representation for RSC prediction.
• Random forest models: Uses random forest classifiers to evaluate cysteine residue labels.
• Different encoding schemes: Applies multiple sequence encoding schemes to represent input protein sequences.
• Sequence-based prediction: Operates solely on protein sequence information to predict redox-sensitive cysteines.
• Probability scoring: Outputs probability scores for each potential RSC residue.
• Performance evaluation: Validated by cross-validation with reported accuracy 0.741 and AUC 0.807 and showing ~10% accuracy and ~13% MCC improvements on independent test datasets.

Scientific Applications:

• RSC identification: Prediction of redox-sensitive cysteine residues in proteins for studies of redox regulation.
• Prioritization for experiments: Prioritizes candidate cysteines for experimental validation in redox biology research.
• Comparative method evaluation: Provides benchmarkable predictions for comparing computational approaches to RSC detection.

Methodology:

Integrates multiple feature representations and applies random forest models with different sequence encoding schemes to generate probability scores for cysteine residues; performance was assessed by cross-validation and independent test datasets reporting accuracy and AUC.