Over the past century, small molecule drugs have represented the dominant modality in drug research, enabling medical breakthroughs that have saved countless lives. A small molecule achieves therapeutic effects by binding to a target protein and blocking or inhibiting the protein’s functions, or changing the protein’s shape in a way that impacts downstream cellular interactions. In either case, the efficacy of a small molecule — also called a ligand — is dependent upon how well it binds to its target protein. This concept, known as protein-ligand binding affinity, is central to small-molecule drug discovery.
Traditionally, small molecule drug discovery teams spend 3-6 years on one program, making and testing anywhere from 1,000-5,000 molecules in a lab. Out of those thousands of molecules, they are hoping to find one with the potency, selectivity, and other desired properties that make it worthy for testing in clinical trials. This process is expensive, time-consuming, and not especially successful. Most programs either fail to identify a satisfactory molecule or only find molecules with suboptimal properties.
At Schrödinger, we are driving a paradigm shift in drug discovery by leveraging digital chemistry. Instead of spending years synthesizing thousands of molecules in a lab, our computational platform allows research teams to explore vast chemical space and predict molecules with optimal properties, so that only the most promising molecules are made and tested in the lab.
One of the technologies underlying our platform is FEP+, an algorithm that uses free energy perturbation calculations — which measure the strength of interactions between two molecules — to predict a small molecule’s binding affinity with its protein partner. Having the ability to accurately and reliably calculate protein-ligand binding affinity is regarded as the holy-grail of computational drug discovery. It allows teams to explore a much larger chemical space and only make and test the molecules predicted to be potent and selective. This dramatically improves the success rate of preclinical drug discovery, leading to higher quality development candidates identified for clinical testing. Multiple studies (from 2017, 2022, and 2023) have demonstrated that the use of FEP+ in active drug discovery programs results in higher quality molecules discovered much faster and with higher success rates compared to traditional trial-and-error methods. Several of the molecules mentioned in the above studies are now advancing through clinical trials.
Our team at Schrödinger has spent many years investing in basic research to develop FEP+, continuously improving its predictive power and expanding its domain of applicability. In a paper published recently in Nature Communications Chemistry, we assembled what is currently the largest and most diverse dataset for the binding of more than 1,200 small molecules to more than 50 proteins. In the study, we conducted 13,732 relative binding free energy calculations to assess the accuracy of FEP+. We found that FEP+ predicted protein-ligand binding affinities with accuracy approaching experimental methods, meaning the tool could confidently and reliably replace the expensive and time consuming trial-and-error methods many small molecule drug discovery teams currently use.
The high level of accuracy and reliability in FEP+ calculations across broad classes of protein targets and diverse chemical series reported in our recent paper represents a major breakthrough in computational drug discovery. As our colleagues at Merck KGaA stated in their recent review of our study, “The experiment is the limit” when it comes to the accuracy of FEP+ and its potential to accelerate innovation in therapeutics. This technology is a game-changer for the discovery of novel small molecules.