bittremieux/tcr-classifier

Classifier to determine CD8+ T-cell receptor characteristics underlying peptide recognition.

tcr-classifier's Issues

Shannon Scores normalisation

Hi,
Dont you think you should have normalised shannon scores to repertoire size? Otherwise, obviously the bigger the repertoire the bigger shannon scores, that seems to be the case in table you have reported int the manuscript.
Thanks

Cannot reproduce accuracy?

I cannot reproduce the accuracy reported for "https://github.com/bittremieux/TCR-Classifier/blob/master/tcr_classifier.ipynb". The only difference with how I run your code is that I do not use joblib. I avoid multi-threading libraries because most python packages are not thread-safe. Rather, I just rerun the code 100 times and average the results. See attachments.
model-rf_hiv.zip

Am I just using your code incorrectly or is there a potential leakage of information across threads?

Thanks
Jared