Hi, I have a very general question and perhaps a bit naive as well. If I have scTCRseq data from 10x, wouldn't it be sufficient for me to run cellranger multi (or cellranger vdj) to obtain information of T cell clone distribution as well as clonotypes of single cells? Does nf-core/airrflow offer anything that is not also offered by cellranger if I only have TCR data? Cellranger performs sequence assembly, mapping, V(D)J annotations, filtering, and outputs clone annotations (airr_rearrangement.tsv) and distributions (clonotypes.csv). Is it your opinion that nf-core/airrflow does it better? Best, Helle

Hi Helle, the airrflow pipeline integrates the functionality of the cellranger vdj tool for exactly these steps that you mention. It then gives out the clonotype tables but also performs further analysis on the cellranger output by integrating it in the immcantation framework of airrflow. There, it performs further filtering, qc, clonal analysis and repertoire analysis and provides you with tables one the one hand but also html reports on the other hand. So, airrflow doesnt do it "better" than cellranger because it uses cellranger. But it expands the functionality while adhering to the FAIR standards.

also to mention that by default, airrflow does also perform V(D)J gene reassignment using IgBLAST, but this is optional and can be omitted with

--reassign false

. We do this step because we require the IMGT gaps to run the downstream tools, and also to use the IMGT reference data, although other reference data can also be provided by the user. Also IgBlast provides the V(D)J allele information in addition to the genes. We also provide clonal diversity and abundance analysis downstream, which are corrected by the sequencing depth of the samples. To say which tool performs best, we would need benchmarks! The advantage of airrflow is that it also deals with bulk data, and mixed bulk and single-cell. This is mostly relevant to the B cell analysis, in case someone else wonders about the same question but for B cells. To define clonal groups (they are called clonotypes by cellranger) we use SCOPER with hierarchical clustering, which is the same implementation in R as Change-O, based in python. This has been shown to perform well compared to other tools in benchmarks by independent groups (e.g. https://www.biorxiv.org/content/10.1101/2024.05.29.596491v1.full.pdf+html, https://bmcimmunol.biomedcentral.com/articles/10.1186/s12865-024-00600-8). The enclone tool that Cellranger uses was not included in the aforementioned benchmarks, it is also closed source so hard to say what is exactly doing. This benchmark does include enclone: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9731466/, which performs worse than SCOPer at 10% mutation rate (a % which is typical for B cell clonal expansion).

Thank you both @Gisela Gabernet and @Mark Polster for your answers. Do you have any publications or tutorials or the like where airrflow or the Immcantation tools have been used on single-cell T cell data?

Hi @Helle Povlsen If you want we can have a short meeting where I can show you the differences from a cellranger output to what else you gain from using airrflow. I must admit though that I am not an expert on Immcantation myself

Hi @Mark Polster, That would be nice! I’m on vacation the next three weeks though, but I’m back July 22nd.