Project Summary Computational chemistry has revolutionized drug discovery, reducing by months or even years the amount of time it takes to discover and refine a lead candidate. While computational chemistry has impacted discovery in many ways, the greatest impact has been via virtual screening (to identify potential hits) and, more recently, through free energy calculations. Free energy calculations such as free energy perturbation (FEP) are used in hit-to-lead improvement. This discovery phase is typically carried out using costly bench chemistry, and tools that can help reduce the number of potential modifications that can appreciably improve efficiency and reduce cost. FEP allows one to evaluate the relative binding of a series of similar ligands to a receptor (protein or nucleic acid) and focus only on those predicted to provide the greatest improvement. While FEP has demonstrated multiple successes in terms of accelerating drug discovery, and while FEP is now integrated as part of drug discovery at most pharmaceutical companies, it suffers from one fundamental weakness, as commonly practiced: FEP is built upon a so-called classical molecular mechanical (MM) energy function. This is a relatively simple function that crudely approximates the real-world energetics. The true energetics are only properly represented using the equations of quantum mechanics (QM). But these equations are so complex, and so computationally difficult to solve, that the entirety of the pharma-relevant toolbox has traditionally integrated the inferior MM approach. FEP has been no different. The result is that while FEP is quite accurate for certain systems where the MM approximations work well, there are a large number of other systems—many of them quite important to drug discovery—for which MM, and MM-based FEP are not reliably predictive. Since discovery projects at pharmaceutical companies are chosen on the basis of biology and not on the basis of their suitability for computational chemistry, it is critically important to identify an approach that can broaden the scope of usefulness for FEP to accommodate more of the targets of interest to pharma. But replacing MM with QM in computational tools has, until our work, led to methods incapable of pharma-relevant turnaround (a few days or less). We have broken this barrier using a hybrid approach QM/MM approach, which provides significantly broader and more accurate coverage of the entirety of the system space of interest to pharma while keeping computational costs and turnaround low enough to be attractive to pharma. Our platform, QUELO, is unique and demonstrably useful. In the first phase of this grant, we implemented, validated, and began marketing it. In this grant phase, we propose algorithmic changes to QUELO to make it more computationally efficient and modifications to allow it to be run on GPU (rather than CPU) computing platforms, which will further reduce total cost and improve turnaround. This will accelera...