A Novel TCR Clustering Method for SARS-COV-2 Epitopes
Abstract
T-cell epitopes are peptides generated from antigens that are presented by MHC class I and class II molecules to T-cells. These epitopes are usually identified by T-cell receptors (TCRs) of CD4 T-cells which then causes transformation of CD4 T-cells to helper or regulatory T cells. In this project, we look at the different TCR sequences of the same length that are activated by two different epitopes for the SARS-COV-2. Building on previous work using Principal component analysis to analyze twenty different physiochemical properties of amino acids, we convert the amino acids in the TCR sequences to numerical strings. We then use four distances methods (Cosine, Cityblock, Euclidean and Correlation) on these strings, and cluster the TCRs in order to compare the dendrogram outputs and see which method does a better job of grouping together like TCRs activated by each epitope. Results are compared to standard matrices such as BLOSUM, PAM, and Gonnet. We thus present a novel TCR clustering technique that will be less computationally strenuous and more cost-effective compared to traditional methods and can be easily utilized by the scientific community to learn more about TCR repertoire sequencing.
Faculty Sponsors
Dr. Radleigh Santos
Project Type
Event
Location
Alvin Sherman Library
Start Date
4-6-2022 12:00 PM
End Date
4-7-2022 5:00 PM
A Novel TCR Clustering Method for SARS-COV-2 Epitopes
Alvin Sherman Library
T-cell epitopes are peptides generated from antigens that are presented by MHC class I and class II molecules to T-cells. These epitopes are usually identified by T-cell receptors (TCRs) of CD4 T-cells which then causes transformation of CD4 T-cells to helper or regulatory T cells. In this project, we look at the different TCR sequences of the same length that are activated by two different epitopes for the SARS-COV-2. Building on previous work using Principal component analysis to analyze twenty different physiochemical properties of amino acids, we convert the amino acids in the TCR sequences to numerical strings. We then use four distances methods (Cosine, Cityblock, Euclidean and Correlation) on these strings, and cluster the TCRs in order to compare the dendrogram outputs and see which method does a better job of grouping together like TCRs activated by each epitope. Results are compared to standard matrices such as BLOSUM, PAM, and Gonnet. We thus present a novel TCR clustering technique that will be less computationally strenuous and more cost-effective compared to traditional methods and can be easily utilized by the scientific community to learn more about TCR repertoire sequencing.
