TCRbase-1.0: a similarity-based model to predict TCR specificity using CDR1, CDR2 and CDR3 loops.
TCRbase-1.0 is a similarity-based model, used to predict TCR specificity. It is solely based on TCR similarities, under the assumption that similar TCRs recognize the same epitope. Kernel similarity [1] similarity measure is used. This measure assigns a similarity score between two sequences by comparing all the k-mers, with k = 1,..,30. For a fixed value of k, the BLOSUM62 score of all the k-mers from the first sequence against the k-mers from the second sequence is computed. The similarity score is then given by the sum of all the BLOSUM scores, for all the values of k.
TCRbase-1.0 requires a training set (database) of TCRs and a test set (query), with the TCRs to predict. Each TCR in the query is scored against the database using the kernel similarity score. The prediction for a given TCR in the test set is then given by the nearest neighbor in the training set. For the CDR3 model, the similarity score is given by the average of similarities of alpha and beta chains. When adding CDR1 and 2 to the model, the overall similarity is given by a weighted average of the similarities of each of the 6 CDR loops (3 for the alpha and 3 for the beta). Previous studies suggest that the CDR3s should be weighted four times higher than CDR1s and 2.
If you need help regarding technical issues (e.g. errors or missing results) contact Technical Support. Please include the name of the service and version (e.g. NetPhos-4.0). If the error occurs after the job has started running, please include the JOB ID (the long code that you see while the job is running).
If you have scientific questions (e.g. how the method works or how to interpret results), contact Correspondence.
Correspondence:
Technical Support: