Contact prediction for beta and alpha-beta proteins using integer linear optimization and its impact on the first...

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Biophysics

Background:

• Protein structure prediction is crucial for understanding protein function.
• Accurate prediction of residue contacts is a key challenge in determining protein tertiary structure.
• Existing methods often struggle with diverse protein folds, including beta, alpha/beta, and alpha+beta structures.

Purpose of the Study:

• To develop an integer linear optimization model for predicting residue contacts in proteins.
• To incorporate hydrophobic interactions and topological constraints into the contact prediction model.
• To evaluate the model's performance and its impact on protein tertiary structure prediction.

Main Methods:

• An integer linear optimization model was formulated to predict residue contacts.
• Protein energy was modeled using C(alpha)-C(alpha) distance-dependent contact energy and hydrophobic contributions.
• A novel hydrophobicity-based method was developed for beta-sheet alignment, used as constraints.
• The model was tested on independent protein datasets and CASP8 proteins.

Main Results:

• The model demonstrated strong performance across beta, alpha+beta, and alpha/beta protein types.
• Average prediction accuracy for contacts separated by at least six residues was approximately 61%.
• Average true positive and false positive distances were 7.58 Å and 15.88 Å, respectively.
• Integration with ASTRO-FOLD improved the quality of predicted protein structure ensembles.

Conclusions:

• The developed residue contact prediction model is effective and accurate.
• The model successfully integrates physical constraints and hydrophobicity for improved predictions.
• This approach significantly aids in protein tertiary structure prediction, as shown by ASTRO-FOLD results.