This repository contains the benchmarking of ZairaChem v1 using the Therapeutics Data Commons Datasets
ZairaChem is an automated pipeline for ML-based (Q)SAR models. Detailed installation instructions can be found in Ersilia's GitBook
In short, to use ZairaChem:
git clone https://github.com/ersilia-os/zaira-chem.git
cd zaira-chem
bash install_script.sh
Model training and prediction:
conda activate zairachem
zairachem fit -i <train_data.csv> -m <model_folder>
zairachem predict -i <test_data.csv> -m <model_folder> -o <pred_folder>
We have benchmarked ZairaChem in the ADMET TDC Leaderboard. At this stage we have focused only on classification tasks.
The admet_classifications notebook shows the code to reproduce the model training and evaluation. For simplicity, the automated reports and raw data of the 5-fold evaluations are provided in the /predictions folder.
| Dataset | Metric | Score |
|---|---|---|
| Bioavailability_Ma | AUROC | 0.74 ± 0.017 |
| HIA_Hou | AUROC | 0.957 ± 0.014 |
| Pgp_Broccatelli | AUROC | 0.944 ± 0.002 |
| BBB_Martins | AUROC | 0.93 ± 0.003 |
| CYP2C9_Veith | AUPRC | 0.792 ± 0.003 |
| CYP2D6_Veith | AUPRC | 0.65 ± 0.09 |
| CYP3A4_Veith | AUPRC | 0.872 ± 0.002 |
| CYP2C9_Substrate_CarbonMangels | AUPRC | 0.421 ± 0.038 |
| CYP2D6_Substrate_CarbonMangels | AUPRC | 0.725 ± 0.006 |
| CYP3A4_Substrate_CarbonMangels | AUPRC | 0.656 ± 0.005 |
| hERG | AUROC | 0.861 ± 0.012 |
| AMES | AUROC | 0.863 ± 0.003 |
| DILI | AUROC | 0.936 ± 0.008 |
If you use our work, please cite us:
The Ersilia Open Source Initiative is a Non Profit Organization with the mission is to equip labs, universities and clinics in LMIC with AI/ML tools for infectious disease research.
Help us achieve our mission!