Developing high-throughput free energy methods for small molecule, peptide, and protein design

The National Institute of General Medical Sciences awarded the University of New Mexico $370,940 on August 22, 2026, under the Biomedical Research and Research Training program (CFDA 93.859) to develop high-throughput free energy computational methods for small molecule, peptide, and protein design in drug discovery and optimization. The project integrates enhanced sampling methods with multisite lambda dynamics (MSLD) to create MSLD++, a framework intended to overcome computational bottlenecks in free energy predictions by enabling rigorous physics-based assessment of thousands of molecular variants in single simulations, particularly for applications involving substantial conformational flexibility in receptors and ligands. Work is performed at the University of New Mexico's Office of Sponsored Projects in Albuquerque, New Mexico, with performance running through May 31, 2031.

Abstract

The discovery and design of therapeutic drugs, including small molecules, peptides, and engineered proteins, are essential for advancing the field of medicine and improving the quality of life. Computational approaches that leverage fundamental physical principles, as well as available structural and functional data of the drug target, can enable more efficient usage of experimental resources, thus reducing the cost and time span of the overall process. Computational tools, such as docking, pharmacophore mapping, and machine learning, can be highly successful in exploring large chemical and protein sequence spaces to identify a small set of lead molecules that show promising activities. Moving beyond the initial discovery stage, rigorous physics-based free energy approaches are often required to more accurately assess and optimize lead molecules. However, a major bottleneck associated with popular free energy methods is the computational cost and thus limited throughput, especially when substantial conformational flexibility exists for the receptor and/or the ligand. To address this bottleneck, we propose to integrate enhanced sampling methods with multisite l dynamics (msld) to achieve an effective computational framework, hereafter referred to as msld++, for rigorous high-throughput free energy predictions in drug discovery and design. Msld enables the modeling of multiple substituents at multiple sites of a common ligand core and can yield free energies for thousands of molecules in a single simulation. IT is thus particularly powerful for solving the combinatorial complexity in the exploration of chemical and protein sequence space frequently involved in lead optimization. Furthermore, this unique feature of msld allows one to run much longer simulations and/or incorporate more sophisticated simulation techniques to adequately sample protein and ligand dynamics for accurate high-throughput free energy calculations, even for challenging cases involving various levels of structural flexibility. The proposed msld++ development will be coupled with three important classes of applications with an increasing level of complexity and demand for conformational sampling: (1) efficient sampling of protein local structural changes in small molecule drug discovery; 2) design of potent and selective therapeutic peptides; (3) understanding and engineering transcription factors for DNA recognition. Successful development of msld++ algorithms will push the limit of accuracy and throughput for free energy- based approaches in the rational design of small-molecule drugs, therapeutic peptides, and functional proteins. Throughout the project, we will engage in close experimental collaborations to critically evaluate the effectiveness of msld++, thus allowing US to deliver robust tools to the broad scientific community.