Back

Predicting epitopes Based on TCR sequence using an embedding deep neural network artificial intelligence approach

Mickael, M. E.; Kubick, N.

2021-08-12 bioinformatics
10.1101/2021.08.11.455918 bioRxiv
Show abstract

T cells receptors are fundamental in recognizing antigens and mediating an appropriate specific immune response against them. Today TCR sequencing has contributed to forming a large repertoire for different immune-associated pathologies. However, predicting epitopes based on TCR sequences has not been satisfactory achieved. We formed a deep neural network using a combination of an embedding autoencoder and selu and relu layers to predict epitopes based on TCR TCR {beta}-chain CDR3. We trained our model using the VDJ database (VDJdb) and validated it using the manually curated catalog of pathology-associated T cell receptor sequences (McPAS-TCR). We used various metrics to measure the accuracy of our tool. We found that our tool can achieve an accuracy of 98 %. Overall our approach presents a step toward identifying microbes crosslinking epitopes that could be playing an important role in various immune diseases.

Matching journals

The top 8 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.