Statistical Learning on Next-Generation Sequencing of T Cell Repertoire Data

• Activity Number: 29 - SPEED: An Ensemble of Advances in Genomics and Genetics
• Type: Contributed
• Date/Time: Sunday, July 29, 2018 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Li Zhang* and Tao He and Alan Paciorek and Jason Cham and David Oh and Lawrence Fong
• Companies: UCSF School of Medicine, UCSF and San Francisco State University and University of California, San Franciscornia and University of California, San Francisco and University of California, San Francisco and University of California, San Francisco
• Keywords: Hierarchical Clustering; Next generation sequencing; Pattern Recognition; Random forest; T cell receptor

Abstract

Immunotherapy has demonstrated a significant clinical activity in cancer. T cells represent a crucial component of the adaptive immune system and are thought to mediate antitumoral immunity. Next generation sequencing of TCR is used as a platform to profile TCR repertoire. We developed an analysis pipeline to track and examine TCR repertoire across time by focusing on V and J gene segments, which overcomes the limitation of small or non-overlap clones among subjects and thus can obtain statistical inferences across subjects directly. We developed a customized clustering workflow: 1) Pattern Recognition by using the combination of V and J gene segments based on their abundance change across time where gap statistics was employed to estimate the optimal number of clusters and k-means algorithm was used for partitioning; 2) Feature Selection to select the important V and J gene segments by random forest; and 3) Hierarchical Clustering to distinguish the subjects based on the selected important V and J gen segments. TCR sequence data from serial peripheral blood mononuclear cells samples from a group of cancer patients receiving the immunotherapy was used for illustration purpose.