TRUST4 config for 10X Genomics

[khoidnyds]

Hello,
I have 2 10X Genomics datasets (BCR and GEX). What are the appropriate parameters setting for TRUST4?
I'm using
f"run-trust4 -1 {f1} -2 {f2} --barcode {f1} --UMI {f1} --readFormat bc:0:15,um:16:25,r1:26:-1 --barcodeWhitelist {whitelist_barcodes} --od {out_dir} -f {hg38_bcrtcr_path} --ref {human_IMGT_path} -t 24 --repseq" for BCR
and
f"run-trust4 -1 {f1} -2 {f2} --barcode {f1} --UMI {f1} --readFormat bc:0:15,um:16:25,r1:26:-1 --barcodeWhitelist {whitelist_barcodes} --od {out_dir} -f {hg38_bcrtcr_path} --ref {human_IMGT_path} -t 24" for GEX
whitelist_barcodes = cellranger-8.0.0/lib/python/cellranger/barcodes/3M-5pgex-jan-2023.txt

but the results are so little of # VDJs compared to Cellranger. I added the first few lines of {f1} and {f2} and index {f1} here too
Thank you

[mourisl]

Which version of TRUST4 are you using? If it is v1.1.1, you can run TRUST4 on the BCR-seq portion without "--repseq" option. How many reads in the toassemble_bc.fq file have "missing_barcode" value? This can be useful to check whether the barcode portion is extracted appropriately. Thank you.

[khoidnyds]

I'm using TRUST4 v1.1.1-r505.
I checked the toassemble_bc.fa and most of the reads having missing_barcode values. Shouldn't the barcodes always be the first 16 bases of R1?
My dataset lengths are mixed. One set has R1 of 26, R2 of 91. Another set has R1 = R2 = 151 bp. What do you recommend for the --readFormat parameter? Thank you

[khoidnyds]

And should I include UMI information when running trust4 or just ignore that?

[mourisl]

The UMI information affects little with the final results, you can include it. Do you see the same fraction of barcode as "missing_barcode" from the GEX side? If you have some of the cellranger VDJ results, you can also check whether the barcode resides on the first 16bp of R1.

[khoidnyds]

Hi mourisl,
Thanks for helping.
I checked the cell_id in _barcode_airr.tsv. Some located at first 16 bp (as expected). but some don't exist in R1.fastq.gz. I can't explain what happened here. Could you give me some hints?
Thanks

[mourisl]

Those could be from error-corrected barcodes.