10x Genomics launches single cell repertoire sequencing


Apparently this was first announced last October at ASHG, but this was news to me:


Seems like 10x knows what they are doing with single-cell sequencing, but hopefully they’ve gotten the rest right.

H/T @bhowie, who might have a comment or two?


@bhowie’s response:


Thanks to @bhowie for his thoughts! Wondering if he (or others) could elaborate on a couple of those points:

In my experience, TCR pairing is neat, but isn’t needed for 99% of rep-seq applications.

Does this imply that T cell diversity can be reasonably estimated by distribution of seqs for a single junction (alpha or beta) or pooled junctions, rather than strictly defining clones based on TCR pairs? My intuition is that ignoring pairing would bias diversity estimates — but I’d be really interested to see any references that suggest otherwise.

It’s really hard to do a great, robust job, esp. w/quantitation.

Curious what level of quantitation you mean here. Because TCR seqs will be directly associated with single cells, it should be straightforward to get absolute and relative numbers of clones in the sequenced population (~1k-10k cells for Chromium, last time I checked). If you actually wanted to quantify the mRNA for TCRs, that’d be more challenging — assuming they need to sequence full length transcripts to reconstruct TCR seqs, that probably means no UMIs.


Does this imply that T cell diversity can be reasonably estimated by distribution of seqs for a single junction (alpha or beta) or pooled junctions, rather than strictly defining clones based on TCR pairs?

Essentially, yes. I’m assuming that one TCRB == one clone, which is overwhelmingly true. If you just want to track clones to see, e.g., how their frequencies change over time or after perturbation, getting paired TCRs costs a lot more money for lower throughput and not much more information.

Basically, my rule of thumb is that if you’re not going to try and make a functional TCR in the lab (or try to computationally model the functions of receptors), you probably don’t need TCR pairs.

Curious what level of quantitation you mean here.

I was trying to make the broad point that sequencing and analyzing VDJ junctions is harder than it seems. Multiplexed primers may not behave; evolutionarily related gene segments create weird biochemical artifacts; low levels of TCR transcripts in T cells hurt efficiency. Basically, it’s a complicated enough system that I set a high bar for believing someone’s FDR for pairing is what they say it is.


Got it. I’ve seen “TCRB == one clone” used as the de facto approach in enough cases, I assumed there was decent logic behind it — just haven’t looked at enough empirical data to really confirm for myself yet. In terms of cost/feasibility, certainly makes sense though. I’ll trust your judgement here.

Good points re: primer behavior, artifacts, etc. I think the “pairing” itself becomes easier when working with single-cell data (though not exactly trivial), but the actual junction sequences being paired might not be as reliable. Will be interesting to see more results from 10x as they become available.



Great, this should help drive costs down for large paired data sets. Has anyone heard the prices involved here?


We just received an announcement that we have a Chromium Machine at the Sanger Institute. I don’t know prces yet, but could easily check.


They will be selling plug and play kits. Had the European rep in here last week but no prices for us yet, nor any ability to answer my questions re methodology. So we wait and see. We decided to go for Dolomite Bio system as is more adaptable for different applications and we have local tech support.


I cannot agree with the “isn’t needed for 99% of rep-seq applications” statement, I’d rather say “isn’t needed for 99% of current rep-seq applications”.


Assuming one nucleotide TCRB sequence is one clone is a good approximation. On the other hand why do we use nucleotide sequences here? Its because different nucleotide assemblies are likely to have different TCR alpha chains.

This is a basic diversity metric, good for counting the number of naive cells, etc. But the real diversity that matters is in the amino acid sequences. For example you can have 10 CDR3beta nucleotide sequences encoding for a single CDR3beta amino acid sequence - this is far less diverse than 10 distinct CDR3beta AA sequences.

If we define the diversity as the number of antigens a given repertoire can efficiently cover, the things get very complex and the amino acid composition of both TCR alpha and TCR beta are needed. For example CDR3AA different by a single amino acid substitution are likely to have an overlapping antigen recognition profile.

Invariant TRA

There are cases when a certain TRA is marking an important T-cell subset, e.g. iNKT cells. You cannot distinguish these cells by simply sequencing TRB. Of course you can do some FACS analysis, but as iNKT is a rare population the contamination will become a problem.

Role of TRA and TRB in antigen recognition

The average number of recognized antigen residues is comparable for TRA and TRB:

So describing tumor-specific TCRs, etc is added to the list


Greetings from the Computational Biology team at 10x.

As you may already know, the Chromium V(D)J T Cell solution is out and users are starting to generate data. In the meantime, we at 10x have made available several datasets using our technology. We hope these datasets will be useful for evaluating the platform and for analysis methods development. You can find the datasets here:


There’s more info on our software solution here:



What about the B cell solution, any news on when it will be released?


We anticipate launching the VDJ B-cell Solution later this year.


Hi Deborah!
I was wondering if you could share any feedback on how the Dolomite system performs in your lab? We are considering to purchase the setup as well, TCR/BCR pairings would be the main goal.


Well firstly I should say it is Alex setting all this up and ironing out the bumps in the lab. He has got the cell/bead encapsulation sorted. Biggest problems encountered were to do with dust blocking the chips which were probably due to the lint free tissue not actually being lint free! He is just finalising the PCR primers and conditions in the lab outside of encapsulation and then will put the two things together. So it wasn’t plug and play, but we didn’t expect it to be and we think it is a more versatile system. Particularly when there are folk around who want to look at different species.


Also I should add that Dolomite have been very supportive


Hi again from 10x Genomics,

We’ve launched our B cell solution along with a 5’ gene expression + V(D)J solution. The latter enables gene expression profiling and paired V(D)J assembly on the same cells.


Public datasets:

Some new datasets showcasing Gene Expression + V(D)J assembly on the same cells:
Gene expression:
T cells:
B cells:

More information on the expression + V(D)J analysis: