This workfow uses a multitask neural network built using PyTorch and saved to the ONNX format to generate predictions. The network generates predictions for bioactivity against 560 targets. The results are visually presented in a bisorted heatmap. The workflow also looks up whatever information is available in ChEMBL itself about the activity of the prediction compounds on the 560 targets and presents that in a second interactive view. NOTE that in order to use this workflow you also need the contents of the Data folder, which contains the saved network and information about the targets that are being predicted.
This is a companion discussion topic for the original entry at https://kni.me/w/ocNbAaAkXf1aQGBQ