MycoMemSVM

MycoMemSVM predicts mycobacterial membrane proteins and classifies their membrane-protein types from primary amino acid sequence to support studies of mycobacterial pathogenicity and drug-target identification.

Key Features:

• Predictive accuracy: Achieves 93.0% overall accuracy for distinguishing mycobacterial membrane versus non-membrane proteins and 93.1% accuracy for classifying membrane protein types.
• Sequence features: Uses over-represented tripeptides extracted from primary protein sequences as predictive features.
• Dataset: Validated on a dataset comprising 295 non-membrane proteins and 274 membrane proteins.
• Validation method: Performance was evaluated using jackknife cross-validation.
• Comparative performance: Demonstrates superior predictive capabilities compared to other existing methods reported by the authors.

Scientific Applications:

• Drug target discovery: Enables identification and classification of membrane proteins to aid discovery of novel drug targets, including targets relevant to multidrug-resistant Mycobacterium strains.
• Functional insights: Provides information on membrane protein types to inform studies of pathogenicity and mycobacterial survival mechanisms.
• Protein annotation: Assists annotation of mycobacterial genome sequences by predicting membrane localization and protein type.

Methodology:

Sequence-based feature extraction using over-represented tripeptides and evaluation by jackknife cross-validation on a dataset of 295 non-membrane and 274 membrane proteins.