An Open Web Service For Fragment-based Design of Small Molecule Inhibitors

Abstract Significance. BMaps™, our fragment-based drug design (FBDD) web application, is a comprehensive tool that empowers chemists in designing small molecules and prioritizing them for chemical synthesis. The transformative impact of FBDD derives from efficiently discovering new chemistries that can exploit the fine 3D details of protein binding surfaces, unlike other molecule screening methods. Further, FBDD enables combinatorial exploration of large chemistry spaces by growing fragments at multiple R-group sites on a molecule. This Phase IIB application has three aims to advance BMaps. By harnessing the power of AI language technologies (e.g. ChatGPT) and incorporating knowledge from building blocks and synthesis schemes from ultra-large chemical libraries such as Enamine and eMolecules, we aim to substantially accelerate drug design workflows and provide tools to generate new ideas for better drug candidates. The design of compounds with optimal affinity, synthetic accessibility, and physicochemical properties is achieved while reducing the workflows from hours to minutes. Innovation. BMaps' core innovation lies in fragment and water maps created from BMaps' GCMC-SACP fragment simulations (Boltzmann maps), which are mined by a geometric search algorithm for optimal compound modifications ranked by binding free energy. This robust framework ensures design efforts that are both expansive in chemistry space and firmly rooted in the principles of statistical mechanics. BMaps will distinguish itself by introducing the first reaction-aware and AI-driven FBDD platform. The proposed extension of our geometric search uses reaction-annotated fragment maps and desired drug properties to radically prune the compound search space compared to affinity-only pruning. These advances of in silico FBDD position our platform to capture a larger market share, boosting sales and user adoption. Approach. First, we will extend chemistry-aware algorithms to dissect chemical building blocks from vendor libraries into simulation-ready fragments, annotated by known reactions. Second, we will incorporate reaction selection and compound filtering interfaces for specifications used in enhancing the efficiency and accuracy of fragment grow searches. Third, we will extend our text and voice AI Large Language Model (LLM) powered tool, Gibbs™, to learn and automate user workflows, and provide expert guidance throughout the design process. Commercialization. By accomplishing these aims, BMaps will have significantly enhanced utility and accessibility that enable the use of FBDD in designing better molecules to a much larger audience. These advancements will bridge the gap to full commercialization, empowering medicinal chemists to efficiently design and prioritize high-affinity, synthesizable compounds with optimal PK properties. BMaps is among the vanguard of the next-generation AI-powered drug design tools.