Building a Virtual Drug Screening Workflow with BioNeMo

Key Takeaways

Nvidia biion Nemo includes an integrated UI and a standalone API that provides seamless access to models and tools that can be used to quickly build custom workflows and Drug Discovery pipelines in this video we'll explore the virtual drug screening workflow provided in the Nvidia biion Nemo GitHub examples to get started we've cloned the GitHub repository and prepared a container with the example notebooks using the included launch script after launching the container we can access the example notebooks in a web browser the virtual screening workflow uses biion Nemo models for protein structure prediction small molecule generation and molecular docking to predict binding Affinity of the small molecule liend and protein Target first we configure access to the API with our NGC API key in the bioo service configuration cell and install a few dependencies we can then check the NGC models available by querying the service to do this we first import Bemo client and then instantiate the python client API using our NGC credentials finding the available models is as simple as calling the list models function in the client API once we've verified access to the service We Begin by looking up the protein sequence for dihydrofolate reductase or dhfr a common binding Target in drug Discovery we then use open fold to predict the 3D protein structure based on its amino acid sequence the next step is generating a set of Li and targets to test docking characteristics for this we'll start with a small database of known dhfr Inhibitors as a seed to the moflow Deep graph generative model to create novel molecules with similar characteristics now that we have our dhfr protein structure from open fold and a set of Target Inhibitors from mofo we can use the diff do docking model to test binding Affinity over a set of poses in this case we'll generate 20 poses for each Li and molecule visualize the docking pose and assess the quality of the results by pose confidence this example workflow provides a starting point for the drug Discovery process using bimo the stages of this workflow can be easily expanded over a broader search space or refined by iterating on high confidence results to generate additional Target molecules by providing access to these models via simple API calls bio allows us to quickly and easily adapt these screening workflows to new use cases