Use Pocket Crafter Online

Pocket Crafter: 3D Generative Chemistry Workflow for Rapid Hit Identification in Drug Discovery

Pocket Crafter is a user-friendly 3D generative modeling workflow for hit finding in drug discovery. It was developed by Lingling Shen, Jian Fang, Lulu Liu, Fei Yang, Jeremy L. Jenkins, Peter S. Kutchukian and He Wang at Novartis Biomedical Research, and published open access in the Journal of Cheminformatics in 2024.

Hit identification is an essential but costly step. High-throughput screening is expensive and time-consuming, which limits the number of targets that can be screened and the diversity of hits for each. Traditional structure-based virtual screening helps, but the authors note there is still considerable room to improve hit rate and chemotype diversity. Many 2D generative models also lack awareness of the 3D shape of the binding pocket.

Pocket Crafter couples the 3D generative model Pocket2Mol, which builds molecules atom by atom inside a protein pocket, with property filters, scoring, structure-activity relationship (SAR) enrichment and clustering to produce a short list of virtual hit scaffolds.

How the Pocket Crafter Workflow Generates and Prioritizes Hit Molecules

1. Define the pocket: start from a tertiary protein structure and the coordinates of the binding pocket centroid.

2. Generate molecules: in the case study, Pocket2Mol sampling was increased to 2,000 and repeated 300 times with different random seeds, giving over 500 thousand de novo compounds.

3. Primary filtering: molecules were validated and kept if molecular weight was at most 800, AlogP between -1 and 7, polar surface area below 125 square angstroms, and fewer than 12 rotatable bonds. This left 352,820 molecules.

4. Hit calling: the remaining molecules were kept as virtual hits if SAscore was 4 or lower, QED was at least 0.5 and the GBVI/WSA dG score was -6 or lower, giving 9,531 virtual hits.

5. SAR enrichment and clustering: a Fisher test identified scaffold clusters enriched among virtual hits (p-value of 0.05 or lower), and the top-scoring molecule in each cluster became a hit scaffold.

6. Library searching: scaffolds were used for shape and electrostatics searching (ROCS) of compound libraries, followed by filters on availability, solubility and quality control.

WDR5 Case Study: Experimental Hit Rate Compared with Library Screening

Library searching produced a focused set of 2,029 compounds, which were tested at 40 micromolar in a WDR5-MYC biochemical HTRF assay. Seven compounds inhibited the signal by more than 40%.

The authors report a 12.8-fold increase in hit rate. Three hits (PC-1, PC-2, PC-3) fitted dose-response curves with IC50 values of 35.6, 27.5 and 28.5 micromolar, were inactive in the counter assay, and PC-1 and PC-2 showed thermal shifts with WDR5 in a differential scanning fluorimetry assay. The hits mapped to chemical space not covered by the earlier screen, and generated molecules reproduced key interactions seen in WDR5 co-crystal structures.

What is Tamarind Bio?

Tamarind Bio is a no-code bioinformatics platform built to give life scientists and researchers access to powerful computational tools. Many cutting-edge machine learning models are hard to deploy and use. Tamarind provides an intuitive, web-based environment that removes the complexity of high-performance computing, software dependencies and command-line interfaces.

The platform is designed for biologists, chemists and other researchers who may not have a background in programming or cloud infrastructure but want to run models on their own data. Key features include:

• A user-friendly graphical interface for setting up and launching experiments

• A robust API for integration into existing research pipelines

• An automated system for managing and scaling computational resources

Tamarind treats information and data security as a top priority, as detailed in its Trust Center and Terms of Service.

Accelerating Discovery with Pocket Crafter on Tamarind Bio

• Hit finding: generate novel scaffolds for a pocket as an alternative or complement to high-throughput screening.

• Protein-protein interaction targets: explore shallow interaction sites such as the WDR5 pocket used in the case study.

• Focused library searches: use virtual hit scaffolds to search archived, commercial or virtual libraries.

• Pocket profiling: examine which chemical groups and interactions the pocket favors, to guide medicinal chemistry.

How to Use Pocket Crafter on Tamarind Bio

1. Log in and open the tool: sign in at tamarind.bio and select Pocket Crafter.

2. Provide the protein structure: upload a 3D structure file of your target, if exposed in that format.

3. Define the pocket: give the binding pocket centroid coordinates, as the workflow requires.

4. Set generation parameters: choose how many molecules to sample if exposed; the paper used large-scale sampling with many random seeds.

5. Run the workflow: submit the job to generate and filter molecules.

6. Download results: retrieve the generated molecules and the filtered virtual hits or scaffold clusters.

7. Select candidates: review top scaffolds and use them for library searching, synthesis or further docking and profiling.

Things to Keep in Mind

• Generated molecules are virtual; hits need experimental confirmation, as in the paper's HTRF and DSF assays.

• The case study hits were weaker than known WDR5 binders WM-662 and Compound 1 (IC50 of 18 and 14 micromolar).

• PC-1 and PC-2 gave negative thermal shifts, which the authors interpret as binding that may destabilize the protein.

• The case study searched Novartis internal libraries, so results depend on the compound sources available.

• Results come from a single target, WDR5.

Source: Shen, L., Fang, J., Liu, L., Yang, F., Jenkins, J. L., Kutchukian, P. S. and Wang, H., "Pocket Crafter: a 3D generative modeling based workflow for the rapid generation of hit molecules in drug discovery," Journal of Cheminformatics 16, 33 (2024), doi:10.1186/s13321-024-00829-w. Pocket2Mol: github.com/pengxingang/Pocket2Mol.