High-throughput sequencing (HTS) has the potential to decipher the diversity of T cell repertoires and their dynamics during immune responses. Applied to T cell subsets such as T effector and T regulatory cells, it should help identify novel biomarkers of diseases. However, given the extreme diversity of TCR repertoires, understanding how the sequencing conditions, including cell numbers, biological and technical sampling and sequencing depth, impact the experimental outcome is critical to proper use of these data. Here, we assessed the representativeness and robustness of TCR repertoire diversity assessment according to experimental conditions. By comparative analyses of experimental datasets and computer simulations, we found that (i) for small samples, the number of clonotypes recovered is often higher than the number of cells per sample, even after removing the singletons; (ii) high-sequencing depth for small samples alters the clonotype distributions, which can be corrected by filtering the datasets using Shannon entropy as a threshold; and (iii) a single sequencing run at high depth does not ensure a good coverage of the clonotype richness in highly polyclonal populations, which can be better covered using multiple sequencing. Altogether, our results warrant better understanding and awareness of the limitation of TCR diversity analyses by HTS and justify the development of novel computational tools for improved modeling of the highly complex nature of TCR repertoires.
Keywords: TCR repertoire; bioinformatics; diversity; normalization; sampling.