Elucidating molecular overlays from pairwise alignments using a genetic algorithm.

Area of Science:

• Computational chemistry
• Molecular modeling
• Drug discovery

Background:

• Inferring relative bioactive poses of active molecules is crucial in drug discovery.
• Pairwise alignment algorithms with rigid conformer libraries are common but limited for multi-ligand overlays.
• Combining pairwise alignments into a single, accurate overlay for multiple active compounds remains challenging.

Purpose of the Study:

• To introduce DIFGAPE (DIstance geometry Focused Genetic Algorithm Pose Evaluator), a novel algorithm for generating accurate molecular overlays.
• To assess DIFGAPE's ability to reproduce observed binding modes from pairwise alignments of active molecules.
• To evaluate the performance of DIFGAPE on diverse test systems using protein-ligand complexes from the Protein Data Bank.

Main Methods:

• Development of the DIFGAPE algorithm, integrating distance geometry and genetic algorithms for pose evaluation.
• Utilizing pairwise alignments of conformers from active compounds as input.
• Testing DIFGAPE on 13 systems across 9 protein targets, starting from 2D ligand structures without protein information.

Main Results:

• DIFGAPE successfully approximated crystallographically observed binding modes in 4 out of 13 test systems.
• For a set of 11 ligands targeting Factor Xa (FXa), the predicted poses achieved a root-mean-square deviation (rmsd) of 1.6 Å compared to crystal structure coordinates.
• The evaluation highlighted existing challenges in molecular conformer generation and pairwise alignment accuracy.

Conclusions:

• DIFGAPE demonstrates potential in predicting ligand binding modes by combining pairwise alignments.
• The algorithm's success in specific cases suggests its utility in drug discovery workflows.
• Further improvements in conformer generation and alignment algorithms are needed for broader application.