TY - JOUR T1 - Related Topics of a Novel TCR-based Cancer Detection Approach JF - bioRxiv DO - 10.1101/2020.07.01.182576 SP - 2020.07.01.182576 AU - Daria Beshnova AU - Bo Li Y1 - 2020/01/01 UR - http://biorxiv.org/content/early/2020/07/02/2020.07.01.182576.abstract N2 - We developed a novel algorithm (DeepCAT) to perform de novo detection of cancer associated TCRs, which is based on a convolutional neural network (CNN) model. In this manuscript, we compared its performance with a similar non-deep learning approach, TCRboost, and demonstrated that DeepCAT achieved better prediction accuracy when used to distinguish cancer from non-cancer individuals. Further, although DeepCAT was trained for CDR3s with different lengths, we showed that the combined outcome does not bias the prediction accuracy. Finally, human immune repertoire is affected by many common inflammatory conditions, and our analysis demonstrated that DeepCAT predictions are minimally affected by these factors.Competing Interest StatementThe authors have declared no competing interest. ER -