This repository contains the code and data for the paper "Evaluation Framework for AI-driven Molecular Design of Multi-target Drugs: Brain Diseases as a Case Study".

The code is written in Python 3.11.6. To install the required packages, run the following command:

$ pip install -r requirements.txt

To download the required data and train the QSAR models for the brain diseases case study, run the following commands:

$ python scripts/external_data_sources.py
$ python scripts/qsar_pipeline.py

The target selction process is described in target-selection/README.md.

1. Build the dataset for the de Novo Design experiment:

$ python guacamol/data/get_data.py --holdout holdout_set_gcm_multitarget.smiles --destination guacamol/data/

2. (Optional) Compute the top K molecules for the de Novo Design experiment:

$ python scripts/top_k.py

The top K molecules are stored in the guacamol/data/top_k folder and save time when running the de Novo Design models over the benchmarks.

3. Run the de Novo Design models over the benchmarks:

$ python scripts/assess_baselines.py

The results of the experiments are saved in the reports folder. We can evaluate the QSAR models on the lead optimization task by running the following command:

$ python scripts/lo_task_benchmark.py