https://github.com/novartis/beyond-pk-score

Prediction of Small-Molecule Developability Using Large-Scale In Silico ADMET Models
https://github.com/novartis/beyond-pk-score

Last synced: 11 months ago
JSON representation

Prediction of Small-Molecule Developability Using Large-Scale In Silico ADMET Models

• Host: GitHub
• URL: https://github.com/novartis/beyond-pk-score
• Owner: Novartis
• License: bsd-3-clause
• Created: 2024-09-05T16:53:30.000Z (almost 2 years ago)
• Default Branch: main
• Last Pushed: 2024-09-08T07:37:17.000Z (almost 2 years ago)
• Last Synced: 2024-09-08T08:20:34.031Z (almost 2 years ago)
• Language: Jupyter Notebook
• Size: 2.48 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# beyond-PK-score

This is the code of the bPK score from the paper 

*Prediction of Small-Molecule Developability Using Large-Scale In Silico ADMET Models* 

by Beckers et al., *Journal of Medicinal Chemistry (2023)*, https://pubs.acs.org/doi/full/10.1021/acs.jmedchem.3c01083

The bPK score is a deep learning based small molecule scoring approach. The model takes as input the predicted ADMET profile of a compound and is trained on historical pre-clinical development milestones using rank-consistent ordinal regression.

The model is based on PyTorch. Training of the model is described in the Jupyter Notebook train.ipynb. The data of public compounds with annotated milestones is provided in public_milestone_dataset.csv.