Pred-binding: large-scale protein-ligand binding affinity prediction.

Area of Science:

• Pharmacology and Drug Discovery
• Computational Chemistry
• Bioinformatics

Background:

• Drug target interactions (DTIs) are vital for drug discovery.
• Experimental determination of compound-protein interactions is resource-intensive and complex.
• In silico methods offer a promising alternative for predicting these interactions.

Purpose of the Study:

• To develop and validate computational models for predicting drug target interactions (DTIs).
• To identify key molecular descriptors influencing ligand-receptor binding affinity.
• To facilitate target discovery and toxicity evaluation in drug development.

Main Methods:

• Utilized Support Vector Machine (SVM) and Random Forest (RF) algorithms.
• Employed 1589 molecular descriptors for compounds and 1080 protein descriptors.
• Trained and validated models on 9948 ligand-protein pairs, quantifying interactions by Ki values.

Main Results:

• Achieved cross-validation coefficients of determination of 0.6079 for SVM and 0.6267 for RF.
• Identified significant descriptors including 2D autocorrelation, topological charge indices, 3D-MoRSE for compounds, and protein autocorrelation and amphiphilic pseudo-amino acid composition.
• Demonstrated the models' efficacy in predicting ligand-receptor interactions.

Conclusions:

• The developed SVM and RF models offer a robust in silico approach for predicting DTIs.
• These models can accelerate drug discovery by predicting ligand-receptor interactions.
• The findings support the use of these computational tools for target identification and toxicity assessment.