Full V(D)J sequence?

[mihem]

Hi @ncborcherding,

for some futher experiment I need the full V(D)J nucleotide sequence. So in 10x output there is fwr1_nt, cdr1_nt, fwr2_nt , cfr2_nt – fwr3_nt – cdr3_nt, fwr4_nt. I think this information gets lost when running combineTCR?
Is this intentionally or just because nobody asked for this yet?

In my use case, I would like to work with the TCR sequenced in the Seurat object (because I've filtered them using scRepertoire, but also scRNAseq information about cell types, so exlucde non T cells) but would then need to have the full VDJ information.

Thank you!
Mischko

[ncborcherding]

@mihem

You are correct - this is actually due to the original Cell Ranger pipeline that did not include the fwr/cdr regions outside of CDR3. So not necessarily intentional and I have not received requests for it.

There is a way around this - you can modify the cdr3 columns in your data to add the additional sequences and combineTCR() will use them - there is similar to the Ibex function here.

modified_data <- lapply(input.data, function(df) {
    if (!all(c("cdr1", "cdr2", "cdr3") %in% colnames(df))) {
      stop("Each data frame must contain 'cdr1', 'cdr2', and 'cdr3' columns.")
    }
    
    # Create concatenated CDR sequence
    df$cdr3 <- paste(df$cdr1, df$cdr2, df$cdr3, sep = "-")
    df$cdr3_nt<- paste(df$cdr1_nt, df$cdr2_nt, df$cdr3_nt, sep = "-")
    
    return(df)
  })

[mihem]

@ncborcherding Thanks for the prompt response.

Ah this makes sense.

Shouldn t this include the fdr region then also? So fwr1_nt, cdr1_nt, fwr2_nt , cfr2_nt, fwr3_nt , cdr3_nt, fwr4_nt?

Wouldn't you think this would be a good default? So that screpertoire by default includes all regions?

Thanks

[ncborcherding]

Hey @mihem

I was giving an example of what you could do to retain sequences - for Ibex - there are models for the cdr3 and models for cdr1/2/3 combined. You can modify the script to include whatever regions you'd like.

In terms of changing the scRepertoire pipeline as a whole - I think it is hard to justify changing as the majority of users have not requested this feature (it is more of an advanced used feature request). I will have to think more on how/if to incorporate larger sequences.

Thanks,
Nick

[mihem]

Yes thanks for this.

Regarding regions: I was thinking as a general solution.

Sure, your decision. I just think more info (in form of more columns) does not have any disadvantages? And why not several columns (as in the 10x output file) instead of one large column?

So my use case is : we need full vdj nt for further experiments. If the Seurat object had not only the cdr3 region but all other regions I could easily combine them with one command.
And if somebody does not need this column they could drop it or not just disregard them?

Thanks,
Mischko

[ncborcherding]

@mihem

I think you make a fair point and I would love to implement this, as well as something similar to mudata in scirpy. The disadvantage is this would require rewriting scRepertoire to handle this - which would be a major overhaul.

I am not sure I can justify that as I have only ever had 2 requests for expanding the sequence storage for scRepertoire. For me, it maybe better to help you modify the pipeline in the short term to work for you and maybe detail the steps in this issue for other users?

In the long term, when I have more bandwidth and major changes to make to scRepertoire, then maybe we can revisit this and use what we learned from your pipeline?

Apologies but hope that might be agreeable.

Nick

[mihem]

Thanks Nick.

Yeah sure, of course you as the maintainer have the best overview what makes sense, how much work this would be and if this might break things.

So what I actually now ended up was really simple: i just read in 10x output and filtered the TCRs I was looking for.
This worked nicely, because we previously identified specific CDR3aa we were interested in.
But if one would want to combine this with data generated by scRepertoire (.e.g. cloneSize) or want this data for all TCR defined by scRepertoire (so some filtered out), it would be handy to have them as additional columns in the Seurat object.

But then let's just keep this open as an enhancement, and if other users or I need this again and you have time, maybe this gets implemented.

Thanks,
Mischko