The Product
Our project addressed the need for a streamlined computational drug discovery process, i.e. the enumeration and generation of chemical structures.
How it works
ProGenix is a command-line interface tool that allows users to:
- Combine chemical building blocks using user-defined reaction rules (SMARTS) to enumerate new molecules
- Validate existing molecules for structural patterns
- Generate new building blocks using a fine-tuned model (DrugEx) targeting specific functional groups
Impact
ProGenix streamlines and automates cheminformatics workflows, which saves time for drug discovery researchers and enables advanced molecular design without programming knowledge.
The Customer
Our client was Remco van den Broek, a PhD candidate at Leiden Academic Centre for Drug Research (LACDR), specializing in computational drug discovery. The communication happened mainly over Microsoft Teams, with both online and in-person meetings.
During the meetings, we had an opportunity to ask for feedback and suggestion that helped with our agile process.
- Use the GitHub Board
The Team
Our Scrum Master was Ernie, and Ryan was the Product Owner. The whole team had software developer roles.
Tasks were divided by functional areas, with a significant amount of work being done through pair programming and frequent group meetings. We used GitHub boards for planning and CI/CD for deployment.
The biggest challenge we've dealt with was our limited knowledge about chemistry and drug discovery. The client helped us with that.
We're most proud of the fine-tuned model that focuses on amines over other groups, as amines were the focus for drug discovery.
The Technologies
- Python
- RDKit
- DrugEx
- Colorama
- Flake8
- Pytest
- RNN (DrugEx Papyrus model)
- SMILES, CSV