MARS: computing three-dimensional alignments for multiple ligands using pairwise similarities.

Area of Science:

• Computational chemistry
• Drug discovery
• cheminformatics

Background:

• Three-dimensional (3D) molecular superimposition is crucial for ligand-based drug design, including quantitative structure-activity relationship (QSAR) and pharmacophore modeling.
• Experimental determination of molecular alignment is often infeasible due to the scarcity of ligand-protein complex structures.

Purpose of the Study:

• To introduce MARS (Multiple Alignments by ROCS-based Similarity), a novel computational approach for achieving accurate 3D molecular alignment.
• To provide a versatile tool for various drug design applications, overcoming limitations of experimental methods.

Main Methods:

• MARS utilizes pairwise alignment of molecules within a dataset using the ROCS (Rapid Overlay of Chemical Structures) tool.
• A score matrix captures pairwise alignment scores, enabling stepwise identification of the optimal superimposition for all molecules.
• The algorithm leverages inter-molecular similarities to compute the best possible 3D alignment.

Main Results:

• The MARS algorithm successfully computes optimal 3D alignments by exploiting similarities across entire datasets.
• Validation through case studies demonstrates the efficacy of MARS for six different datasets.
• The tool supports diverse applications including pharmacophore generation, 3D QSAR, clustering, outlier identification, and alignment expansion.

Conclusions:

• MARS offers a robust computational solution for 3D molecular alignment, essential for advancing ligand-based drug design.
• The method effectively addresses the challenge of limited experimental structural data.
• MARS is a valuable tool for multiple cheminformatics and drug discovery tasks.