BCov

BCov predicts β-sheet topology from protein sequence and secondary structure to identify β-strand pairings and overall β-sheet connectivity.

Key Features:

• Input data: Uses protein sequence and predicted or assigned secondary structure as the starting information for topology prediction.
• Unsupervised methodology: Predicts β-strand contacts and connectivity without relying on supervised training data specific to β-sheet topology.
• Sparse inverse covariance estimation: Computes β-strand partner scores by estimating sparse inverse covariance to identify residue-residue statistical couplings, including signals from correlated mutations and sequence-derived features.
• Integer programming optimization: Applies integer programming to score-derived predictions to determine feasible and optimal β-sheet connectivity.
• Performance on atomic-resolution dataset: Evaluated on a non-redundant set of 916 protein chains with atomic resolution, achieving average Matthews Correlation Coefficient (MCC) of 0.56 and F1 score of 0.61 for β-strand pairing prediction.
• Comparison with state-of-the-art methods: Produces results competitive with existing state-of-the-art methods trained for β-sheet topology prediction.

Scientific Applications:

• Protein structure prediction: Supports modeling of three-dimensional protein structures by predicting β-strand connectivity and pairing.
• Functional and interaction inference: Provides structural constraints that aid in understanding protein function and intermolecular interactions.
• Structure-guided drug design: Supplies β-sheet topology information that can inform therapeutic design efforts dependent on protein structure.

Methodology:

BCov computes β-strand partner scores using sparse inverse covariance estimation from sequence and secondary-structure-derived signals, then applies an integer programming algorithm to predict the most plausible β-sheet connectivity based on those scores.