Fragment-similarity-based QSAR (FS-QSAR) algorithm for ligand biological activity predictions

Area of Science:

• Computational chemistry
• Medicinal chemistry
• Drug discovery

Background:

• Quantitative structure-activity relationship (QSAR) studies are essential computational tools in drug discovery.
• Fragment-based drug design (FBDD) is a key strategy for developing novel therapeutics.
• Integrating FBDD concepts with QSAR can enhance predictive modeling.

Purpose of the Study:

• To develop a robust fragment-similarity-based QSAR (FS-QSAR) algorithm.
• To correlate chemical structures with biological activities using FBDD principles and multiple linear regression.
• To validate the predictive performance of the developed FS-QSAR method.

Main Methods:

• Developed a novel FS-QSAR algorithm integrating FBDD and multiple linear regression.
• Calculated fragment similarity using eigenvalues of BCUT matrices and Tanimoto coefficient (Tc) of molecular fingerprints.
• Applied the algorithm to datasets of COX2 inhibitors and cannabinoid CB2 receptor antagonists for model building and validation.

Main Results:

• The FS-QSAR models achieved predictive accuracies with coefficients of determination (r²) of 0.62 for COX2 inhibitors and 0.68 for CB2 antagonists.
• The developed method demonstrated superior prediction accuracy compared to traditional and one-nearest-neighbour QSAR approaches.
• BCUT-similarity and Tc-similarity functions proved effective in correlating fragment structures with biological activity.

Conclusions:

• The developed FS-QSAR method provides accurate ligand activity predictions.
• This approach is a valuable tool for fragment-based drug discovery.
• FS-QSAR enhances the efficiency of identifying potential drug candidates by predicting ligand activity.