Fig. 1.

Workflow diagram of DeepNetBim framework. First, the binding and immunogenic data were retrieved from the IEDB database. Then, the weighted HLA-peptide binding network (coloured in blue) and immunogenic network (coloured in purple) were constructed separately to acquire quantified network features. Next, the encoded peptides and the obtained network features were fed into an attention-based deep learning process. Finally, the predicted binding affinity and binary immunogenic category of the above two independent models were combined to make the final prediction. BA binding affinity, IC immunogenic category