About

PermMol is a pretrained model having ~87,000,000 parameters. PermMol is based on XLNet and features pretraining mechanism on tokens' permutation, two-stream self-attention and molecule graph attention, supporting multiple and accurate ADMET properties prediction. With AI booms in different scopes, predictive method is becoming more and more effective and efficient in medicinal research. We believe this powerful tool can boost drug discovery to a large extent.

If you would like to contact us for collaboration, please make connections with Dr. Lin. If you have any questions or meet any problems about this project, please email developers. It's very appreciate receiving your feedback.