Significance: To date, no specific therapeutic drug or vaccine has been approved for the treatment of human coronavirus. Better, direct?acting anti?viral drugs and accelerated methods for identifying them are desperately needed. Having a large body of diverse fragment binding simulation data for each CoV2 drug target represents a unique opportunity to accelerate preclinical drug discovery for CoV2 protein inhibitors. In contrast to testing?based approaches, understanding fragment interaction patterns provides chemists specific mechanistic information to guide lead optimization. We propose to (1) create comprehensive fragment maps for the full suite of CoV2 proteins; (2) build automated tools for enumeration and evaluation of compounds that address protease selectivity and inhibition at Spike protein ppi and allosteric sites; and (3) make these available worldwide through the BMaps Web application. As such, all anti?viral researchers can benefit. Innovation: Generating thousands of fragment binding patterns for each of the known CoV2 protein structures is a novel scientific approach to the rational design of CoV2 antivirals. This would be the largest data source of fragment data on CoV2 drug targets available and the resource would be accessible by all scientists working to address the COVID?19 pandemic. The innovation proposed is to enable a new scientific approach to rational design for CoV2 antivirals based on the analysis of fragment binding patterns using novel compound enumeration and evaluation methods.
Aim 1 : Generate fragment and water maps for the full suite of proteins involved in the coronavirus life cycle. Using hot spots for location bias, run ~1,000 fragment simulations on each consensus of 6 structures from molecular dynamics.
Aim 2 : Develop automated tools to accelerate the enumeration and evaluation of candidate inhibitor molecules. Two approaches are proposed: (1) adapt our test software to enumerate all available modifications with all fragments for a given starting point and (2) use a Conditional GAN (Generative Adversarial Network) deep learning network to enumerate inhibitors from fragments, using discriminator networks to bias towards synthesizable molecules with good properties.
Aim 3. Build a repository of candidate inhibitors targeting coronavirus proteins through a variety of different mechanisms. Overall Impact: The CoV2 protein?fragment maps lead chemists to often non?obvious ideas to progress their compounds toward clinical trials. The ability to automatically enumerate and evaluate compounds from a large fragment map repository enables broad access to target?relevant chemical diversity, without tedious manual searching. A repository of candidate inhibitors targeting coronavirus proteins enables drug researchers to get started quickly.
Accessing diverse chemical fragment binding data (1,000s for each CoV2 drug target) presents a unique opportunity to accelerate much needed preclinical drug discovery for CoV2 inhibitors. The fragment-based drug design platform (BMaps, www.boltzmannmaps.com), built under NIH Phase I and II grants, will be extended with fragment maps for CoV2 proteins and further automation of compound enumeration through fragment growing. This will empower antiviral drug researchers in attacking the difficult problems of CoV2 protease selectivity and inhibitor binding at allosteric or protein-protein interaction sites on the CoV2 S protein.
