Ian D. Wall

Alchemical free energy (AFE) calculations are a useful tool in computational drug discovery. However, they typically involve relatively short (<10 ns) simulations, meaning that the initial coordinates and, more generally, the setup of the system have a significant effect on the obtained free energy values. To remedy this, we recently developed a fully adaptive version of the simulated tempering algorithm (FAST) and applied it in the context of sampling. In this work, we extend FAST to AFE calculations with and without enhanced sampling of a particular degree of freedom of interest (FAST/MBAR). We show that enhanced sampling significantly increases the mobility of the targeted degree of freedom at the cost of reduced sampling efficiency over λ space. On the other hand, the free energy calculations without explicit targeting of certain degrees of freedom retain initial-coordinate bias over longer time scales. Despite this, both protocols readily explore nanosecond-time-scale events, such as torsional rotation, due to the single-trajectory nature of FAST, making them less sensitive to the system preparation. It is shown that the robust automated nature of FAST/MBAR makes it a competitive alternative to conventional AFE methods.

Endoplasmic reticulum aminopeptidase 1 (ERAP1) regulates immune responses by proteolytically processing peptides presented by major histocompatibility class I molecules (MHC-I). ERAP1 can reduce the immunogenicity of cancer cells by destroying cancer-associated antigenic peptides or contribute to autoimmunity by generating self-antigenic peptides. ERAP1 inhibition has emerged as a tractable approach for cancer immunotherapy and specific classes of autoimmune diseases. We describe the discovery of a potent and selective ERAP1 inhibitor that targets its regulatory allosteric site. The compound has favorable in vivo pharmacokinetics, oral bioavailability, can regulate the immunopeptidome of cancer cells, and enhance tumor antigenicity in vivo controlling growth. When administered in the murine collagen-induced arthritis model, we observed no exacerbation of autoimmune responses but rather a dose-dependent therapeutic benefit. Our results demonstrate that ERAP1 inhibition is a tractable approach to modulating immune responses, provide mechanistic insight, and are valuable in vivo tools for interrogating ERAP1 biology and further drug development.

Generative design and machine learning are increasingly prevalent in medicinal chemistry. To pilot the comprehensive use of automated molecular design on a project, the BRADSHAW platform was used to optimize a series of inhibitors of Endoplasmic Reticulum Aminopeptidase 1 (ERAP1), an emerging target in cancer immunotherapy and autoimmune diseases. Through four consecutive iterations applying in silico molecular generation, property prediction and filtering, we conducted a multiparameter optimization of potency, physicochemical properties and pharmacokinetics. Continuous refinement of Machine Learning (ML) models led to improved scoring accuracy and compound quality, culminating in identification of in vitro and in vivo tool molecules. We also discuss our reflections on the pilot and integration of automated design into medicinal chemistry projects, including observations of the human factors resulting from increased use of computational design, and recommendations for future projects.