VICMpred

VICMpred predicts major functions of gram-negative bacterial proteins from amino acid sequences by classifying them into virulence factors, information molecules, cellular process molecules, and metabolism molecules.

Key Features:

• Function classification: Classifies gram-negative bacterial proteins into four categories: virulence factors, information molecules, cellular process molecules, and metabolism molecules.
• Machine learning: Employs Support Vector Machines (SVM) as the primary classification algorithm.
• Input features: Uses amino acid composition, dipeptide composition, tetrapeptide-derived features, and a hybrid of tetrapeptide plus amino acid composition as model inputs.
• Tetrapeptide selection: Identifies unique tetrapeptides significantly associated with each protein class for use as features.
• Training dataset: Trained and tested on 670 non-redundant gram-negative bacterial proteins comprising 255 cellular process proteins, 60 information molecules, 285 metabolism-related proteins, and 70 virulence factors.
• Performance metrics: Achieved overall accuracies of 52.39% (amino acid composition), 47.01% (dipeptide composition), 68.66% (tetrapeptide method), and 70.75% (hybrid tetrapeptide plus amino acid composition).
• Cross-validation: Model performance was assessed using five-fold cross-validation.

Scientific Applications:

• Functional annotation: Supports sequence-based annotation of gram-negative bacterial proteins into major functional categories.
• Pathogenesis and cellular biology studies: Aids investigation of bacterial virulence, cellular processes, and metabolic functions from protein sequences.
• Therapeutic research support: Provides functional predictions that can inform development of strategies targeting gram-negative bacterial infections.

Methodology:

Support Vector Machine models were trained and tested on a dataset of 670 non-redundant gram-negative bacterial proteins (255 cellular process, 60 information molecules, 285 metabolism, 70 virulence) using amino acid composition, dipeptide composition, tetrapeptide-derived unique tetrapeptides, and a hybrid of tetrapeptide plus amino acid composition, with evaluation by five-fold cross-validation.