A transferable multi-channel model for detecting β-lactamase - inhibitor interaction
Python3, PyTorch, RDkit, DeepPurpose
The label reversal KIBA dataset can be downloaded from TransformerCPI.
Beta-lactamase dataset we collected can be found in this repository.
DeepBLI is developed based on DeepPurpose library. Modification for multi-channel model are in DTI_multi.py
and utils_multi.py
files. So first please copy these two files to yourpath/DeepPurpose.
run.py
involves the pre-training, fine-tuning and predicting parts of code. We also provide the trained DeepBLI model.