USING MACHINE LEARNING TO STUDY TCR α β CHAIN PAIRING: CLASSIFICATION APPROACH

Abstract

Paired TCRα β chains play a critical role in the immune system in which their main responsibility is to host an immune response against pathogens. However, recognising the paired and unpaired TCR chains is still an open problem. This project tackles this is- sue by studying the TCR chain pairing by following a classification approach using machine learning algorithms. To achieve this goal, two large paired and unpaired TCR datasets are analysed, and pre- processed using integer encoding and one-hot encoding in Python. Multiple experiments are conducted to predict if a given TCR sequence is paired or not using supervised classification algorithms, ensemble methods, and artificial neural networks. Further, the classification models are optimised and compared using various evaluation metrics.