TBBpred

TBBpred predicts transmembrane beta-barrel regions in membrane protein sequences to identify structural elements involved in membrane transport and enzymatic functions.

Key Features:

• Prediction target: Predicts transmembrane beta-barrel regions in membrane proteins.
• Artificial Neural Network (ANN): Uses a feed-forward ANN trained by back-propagation that incorporates evolutionary information from multiple sequence alignments obtained via PSI-BLAST.
• Support Vector Machine (SVM): Employs an SVM that inputs primary protein sequences augmented with 36 physicochemical parameters.
• Combined ANN+SVM fusion: Integrates ANN and SVM methodologies to improve predictive performance, achieving 81.8% accuracy and a Matthews correlation coefficient (MCC) of 0.64.
• Validation dataset and protocol: Evaluated on a nonredundant dataset of 16 proteins using leave-one-out cross-validation (LOOCV).

Scientific Applications:

• Protein annotation: Assists annotation of newly sequenced proteins by identifying transmembrane beta-barrel regions.
• Structural modeling: Provides predicted beta-barrel region locations to support membrane protein structural modeling.
• Functional studies: Supports studies of membrane-bound enzymes and transporters by locating transmembrane beta-barrel elements relevant to function.
• Membrane protein research: Aids research into membrane protein structure and function through targeted beta-barrel predictions.

Methodology:

Combines a feed-forward ANN trained by back-propagation with evolutionary information from PSI-BLAST multiple sequence alignments, an SVM using primary sequences plus 36 physicochemical parameters, and a fusion of ANN and SVM predictions; performance was assessed by leave-one-out cross-validation on a nonredundant set of 16 proteins.