Estimation of TCRdiv

[bravib]

Good evening,
after reading your paper I would be interested in estimating TCRdiv of a series of epitope-specific repertoires, however I do not find a script dedicated only to it without the need of running the full analysis. Could you please help me with this? Does one need also the alpha-chain to estimate TCRdiv?
Many thanks, best regards,
Barbara

[phbradley]

Hi Barbara,
Thanks for your query. Do you have paired (alpha/beta) data or just single chain?

[bravib]

I actually have only beta chains, so I was wondering if the TCRdiv measure is applicable, limiting the evaluation of the distance to these chains.
Many thanks,
Barbara

[phbradley]

OK, one more question-- roughly how big are your repertoires? Hundreds, thousands, millions? I also have a C++ repository that can compute TCRdiv and scale to larger datasets...

[bravib]

I see, in terms of unique sequences they are between 25000 and 50000 and I look at the CDR3 region (while I know that TCRdist is calibrated not only on the CDR3). Thanks a lot for your help,
Barbara

[phbradley]

OK, and do you only have the CDR3 or do you mean that when you look only at the CDR3 you have 25000 to 50000 unique sequences?

[bravib]

Ah sorry, I mean that I have only the CDR3 and they are roughly between 25000 and 50000 unique ones.

[phbradley]

Hi Barbara, without the V genes it's not really possible to compute TCRdist. You could try something like a simple Hamming distance or Levenshtein distance on the CDR3s and try using the same formula for TCRdiv, but I don't have code to help with that. Sorry!

[bravib]

Yes, I was already thinking about proceeding in this way, many thanks for your replies.
Barbara