powerTCR: A model-based approach to comparative analysis of the clone size distribution of the T cell receptor repertoire

• DOI: 10.1371/journal.pcbi.1006571
• Publication Date: 2018-11-28T18:27:41Z
• Authors: Hillary Koch; Dmytro Starenki; Sara J. Cooper; Richard M. Myers; Qunhua Li
• Keywords: CD4-Positive T-Lymphocytes; 0301 basic medicine; Likelihood Functions; Sarcoidosis; QH301-705.5; Brain Neoplasms; Receptors, Antigen, T-Cell; High-Throughput Nucleotide Sequencing; Clone Cells; 3. Good health; Alternative Splicing; Mice; 03 medical and health sciences; Immune System; Animals; Cluster Analysis; Humans; Computer Simulation; Programming Languages; Biology (General); Glioblastoma; Lung; Software; Research Article

AbstractSequencing of the T cell receptor repertoire is a powerful tool for deeper study of immune response, but the unique structure of this type of data makes its meaningful quantification challenging. We introduce a new method, the Gamma-GPD spliced threshold model, to address this difficulty. This biologically interpretable model captures the distribution of the TCR repertoire, demonstrates stability across varying sequencing depths, and permits comparative analysis across any number of sampled individuals. We apply our method to several datasets and obtain insights regarding the differentiating features in the T cell receptor repertoire among sampled individuals across conditions. We have implemented our method in the open-source R package powerTCR.Author summaryA more detailed understanding of the immune response can unlock critical information concerning diagnosis and treatment of disease. Here, in particular, we study T cells through T cell receptor sequencing, as T cells play a vital role in immune response. One important feature of T cell receptor sequencing data is the frequencies of each receptor in a given sample. These frequencies harbor global information about the landscape of the immune response. We introduce a flexible method that extracts this information by modeling the distribution of these frequencies, and show that it can be used to quantify differences in samples from individuals of different biological conditions.