Code repository for manuscript: TransMHCII: a novel MHC-II binding prediction model built using a protein language model and an image classifier
- input_with_features.csv: Same as the Supp.Table 1, but formatted in CSV.
- embed.ipynb: example embedding using PLM model
prot_t5_xl_uniref50
. Inputs the CSV file above, and outputs a pickle file. Pickle file is not uploaded to this repository due to size. - train_ProtT5_efficientnet.py: example code for constructing and training the ProtT5 EfficinetNet v2b0 model. Inputs the CSV file and pickle file from above. A tensorflow model is built in the process. We used a g4dn.12xlarge instance on AWS to train the model in ~ 4 hours.