Predicting the binding residues of four small molecule ligands by utilizing ensemble algorithms with additional...

Area of Science:

• Computational Biology
• Bioinformatics
• Drug Discovery

Background:

• Accurate prediction of protein-small molecule ligand binding residues is essential for protein functional annotation and molecular drug design.
• Existing sequence-based prediction methods have limitations in improving prediction accuracy.
• Precisely identifying binding residues for ligands such as ATP, ADP, GDP, and NAD remains a challenge.

Purpose of the Study:

• To develop a novel computational method for precisely identifying protein-small molecule binding residues.
• To introduce and evaluate four new correlation features based on sequence information.
• To enhance the prediction accuracy of binding residues for ATP, ADP, GDP, and NAD ligands.

Main Methods:

• Developed a prediction method incorporating four correlation features: neighbor correlation, residue pairs, central motifs, and PSSM correlation.
• Utilized both a self-built dataset and datasets from previous studies for validation.
• Conducted ablative experiments to assess the contribution of each correlation feature.

Main Results:

• The proposed method achieved favorable prediction results on independent testing datasets.
• Highest performance metrics were observed for specific ligands: Sensitivity for GDP (54.93%), Specificity for ATP (98.68%), Accuracy for NAD (53.41%), and MCC for NAD (0.5341).
• Ablative studies confirmed that the introduced correlation features significantly enhance prediction performance.

Conclusions:

• The novel correlation features significantly improve the prediction of protein-small molecule binding residues.
• The developed method offers a robust approach for identifying binding sites, aiding in functional annotation and drug design.
• The selected feature parameters and algorithms are crucial for building effective prediction models.