search

lgragert / nn-sero-pytorch

PyTorch version of neural network HLA serology prediction
2 stars 1 forks source link

Inconsistencies between new and old polymorhphic AAs #31

Closed gbiagini closed 4 years ago

gbiagini commented 4 years ago

It seems that, in even the "complete" versions of the training and testing sets, which are meant to include basically all possibly polymorphic residues, there are some possible values from the original data that are not included in the current datasets.

gbiagini commented 4 years ago

This seems to have been fixed by including the alleles from the training and validation in the testing set.