Image created with Midjourney and modified by Jana Schor, 2022.
AI on the Chemical Universe
We develop AI tools like deepFPlearn to predict the associations between chemicals & effects for 100s of 1000s of compounds in seconds. By including explainable AI and Quantification of Uncertainty, we enhance the credibility of the tool’s predictions for any set of compounds.
Goal: Associate chemicals, represented by their structures, with biological effects
In this project we are working with:
- Feed forward neural networks using Morgan Chemical Fingerprint representations for chemicals as input
- Graph neural networks with graphs to represent chemical structures
- Autoencoder networks to pretrain the feed forward neural networks
- Explainable AI and Methods to quantify uncertainty of computational predictions
Related publications
- Soulios, K., Scheibe, P., Bernt, M., Hackermüller, J., Schor, J. (2023):
deepFPlearn+: enhancing toxicity prediction across the chemical universe using graph neural networks
Bioinformatics 39 (12), btad713 10.1093/bioinformatics/btad713 - Soulios, K., Scheibe, P., Bernt, M., Hackermüller, J., Schor, J. (2023):
deepFPlearn+
Zenodo 10.5281/zenodo.8146252 - Schor, J., Scheibe, P., Bernt, M., Busch, W., Lai, C., Hackermüller, J. (2022):
AI for predicting chemical-effect associations at the chemical universe level — deepFPlearn
Brief. Bioinform. 23 (5), bbac257 10.1093/bib/bbac257