How to retrieve back Sequence ID from the successful aligned .vdjca file?

[ShaowenJ]

Dear MixCR developers,

I have a quick question, do you know how to or in what parameters can I retrieve back the fastq Sequence ID in the alignment step? I know you can get the aligned Sequence by "mixcr exportAlignments", but don't see the sequence ID.
I am using the Nanopore preset.
If there's one, that will be great! Otherwise I could search by the aligned sequence.

Best!

[mizraelson]

Hi,
Yes, it is possible to export both read Id and a description from the original FASTQ file(to do that you have to add a parameter in mixcr align):

mixcr align \
    --species hsa \
    -OsaveOriginalReads= true \
   .
   .
   .

Then in mixcr exportAlignments you can add columns with the reads description from the original fastq file and read Ids:

mixcr exportAlignments \ 
    -descrsR1 \
    -descrsR2 \
    -readIds \
   .
   .
   .

-descrR1 and -descrR2 add fastq original headers for every read, while -readIds adds a unique ID number assigned by MiXCR. For read Id there is no need to save original reads but you can't easily backtrack alignments to the original reads by that id.

Also, if you want you can add -OsaveOriginalReads= true parameter to the mixcr analyze command so you don't have to rerun the alignment:

mixcr analyze ont-rna-seq-vdj-full-length \
     --species hsa \
     -Malign.parameters.saveOriginalReads=true \
     sample.fastq.gz \
     sample_result

But do keep in mind that this parameter increases the size of .vdjca file because it attaches all the original reads.

[ShaowenJ]

Hi @mizraelson, thanks for the information!
I checked the overlap of -cloneId with the union of all cloneIDs from the tsv files, they are totally overlapping except for -1. Is this mean they are not reassigned?

[ShaowenJ]

Hi @mizraelson I would share the structure of our library here from the other discussion.
Basically, each sequence should be composed as this diagram:
Barcode(14bp)+UMI(7bp)+VV(V: any nucleotide but T)+polyA+cDNA
The Barcode should be matched from a whitelist file like a single-cell method. I can separate by myself.
Currently, I was using PolyA+cDNA as bulk sequences to run the MiXCR ONT preset and trace back the barcode and UMI information by sequence id (I have a dataframe keep this information). But I see a higher number of clones than the number of UMI for most barcodes. The reason for this I guess is the error rate for longer sequences like cDNA is much higher than the UMI, which generates artifact clones caused by sequence errors. So I should perform some UMI demultiplexing to combine these same or similar UMI.
Do you have any suggestions on how to perform it from MiXCR? That would be great!

Any ideas would be appreciated!

Thanks!

[mizraelson]

So if I understand correctly th first 14bp is the cell barcode? can you share the whitelist for that?
VV - two nucleotides (anything but T)
Does the length of the polyA vary or is it always the same?
Then you have the full receptor sequence including 5'UTR?

Also, is it possible to share some test raw file (with no preprocessing) so i can test the preset?

[ShaowenJ]

Hi @mizraelson , yes sure. How can I share these with you guys? I prefer by email.
The polyA varies, but I guess I could remove the polyA from alignment and only put the cDNA.
Theoretically, we hope to recover the full sequence, but in reality, it’s not always the case for our results.

Thanks

[mizraelson]

Sure, you can use the following email.

[email protected]

Also if the test files are large we have a dedicated server for uploading it.