# Develop

Develop is a deep generative design model that proposes novel, drug-like molecules by applying 3D pharmacophoric constraints. It is open source and self-hosted, installable on Docker Hub.

Canonical page: https://ourlocalhost.com/app/5765/develop

## Description

DEVELOP is a deep generative design model that proposes novel, drug-like molecules by applying 3D pharmacophoric constraints. It is the open-source reference implementation of the 'Deep Generative Design with 3D Pharmacophoric Constraints' method and ships as a self-hosted Docker container, letting computational chemistry and drug-discovery researchers generate structurally valid, target-relevant candidate compounds on their own server.

## Platform availability

Available on 0 of 10 platforms.

## Health score

26/100 (Low), based on 5 of 5 factors.
- Maintained: 0 (last commit 1774 days ago)
- Popular: 24 (14 GitHub stars)
- Easy to install: 25 (0 install platforms, install notes, Docker image)
- Light to run: 20 (16 GB minimum RAM, runs on ARM)
- Security check: 85 (no serious findings, tests)

## Links

- GitHub: https://github.com/fimrie/DEVELOP
- Docker Hub: https://hub.docker.com/r/chaste/develop

## Common questions

### How do I set up the DEVELOP environment?

Create it with conda from the provided DEVELOP_env.yml file, then activate DEVELOP_env. The code was tested with Python 3.7, Tensorflow 1.14, and molgrid 0.2.1.

### Which data-preparation scripts should I run for the two main workflows?

Run prepare_data_linker_design.py for linker design and prepare_data_scaffold_elaboration.py for scaffold elaboration from the data directory. You can pass options such as --output_dir, --add_idx, and --include_pharms.

### How do I train and generate molecules for linker design?

Run DEVELOP.py with --dataset zinc and a config JSON specifying train and valid molecule files, struct files, epoch settings, and the struct data root. Scaffold elaboration can be performed analogously.

### How can I generate molecules with a target atom count instead of matching the reference molecule?

Use the generation setting that produces elaborations with a specified number of atoms. The other setting generates molecules with the same number of atoms as the reference molecule.

---
Source: Our Local Host (https://ourlocalhost.com/app/5765/develop), the self-hosted and open-source app directory.
