Spl-IsoQuant with 10x 3' v4 single-cell

[Suhail221107]

Hello IsoQuant team,

Thank you for developing this tool. I am planning to analyze our single-cell long-read PacBio data using spl-IsoQuant. While reviewing the supported protocols, I noticed that it currently lists support for the 10x Genomics 3′ v3 single-cell protocol. However, our samples were processed using the 10x 3′ v4 chemistry. Would the pipeline still work if I provide the corresponding whitelist for the 10x 3′ v4 protocol, or is it limited to 10x 3′ v3 only?

Best regards,
Suhail

[andrewprzh]

Dear @Suhail221107

As far as I understand, the molecule structure for v4 is identical to v3, so providing the correct whitelist will do the job.

I also recommend to use a shorter whitelist that contains only captured cells (if you can get one using short-read sequencing) instead of the full whitelist. This dramatically increases barcode detection accuracy.

Also, current release is rather slow with processing large whitelists for 10x data. I will make a new release within next few days. It should work much faster.

Best
Andrey

[Suhail221107]

Dear Andrey,

Thank you very much for your quick response. I have a follow-up question: can I provide barcode-corrected and deduplicated BAM files produced by the Iso-Seq tools as input in cases where I want to skip barcode correction? Or is it better to perform barcode correction and deduplication within IsoQuant instead? What would you suggest?

Best regards,
Suhail

[andrewprzh]

@Suhail221107

At the moment IsoQuant cannot read third-party barcodes and UMIs from a BAM file - this will be also implemented in the near future.
However, you can extract read ids and corresponding barcode and UMI tags from a BAM file into a TSV with 3 columns and feed it into IsoQuant via --barcoded_reads option.
In general, I don't see any problem in using third-party barcodes.

Also, if you want to get barcode-grouped counts, you have to specify that explicitly via --read_group barcodes.

Best
Andrey

[Suhail221107]

Dear Andrey,

Thank you for the clarification; it is very helpful.

Best,
Suhail

[Suhail221107]

Hi Andrey,
I recently ran the pipeline on our single-cell PacBio data using the deduplicated BAM files from Iso-Seq and also provided the --barcode_reads option. Below I’ve included the full script and corresponding output. I’m unsure which file from these results should be used as input for Seurat or Scanpy, since I don’t see any output with an .mtx suffix. Do I need to run an additional step to generate a count matrix file that is compatible with Seurat/Scanpy. Could you please guide me on this issue.

Best,
Suhail

python splisoquant.py -d pacbio_ccs
--unmapped_bam sample1_dedup.bam sample2_dedup.bam sample3_dedup.bam sample5_dedup.bam sample5_dedup.bam sample6_dedup.bam sample7_dedup.bam sample8_dedup.bam sample9_dedup.bam
--reference genome.fa --genedb genes.gtf --complete_genedb --output pacbio_scIsoquant --prefix islet_pacbio_n9 --polya_trimmed all
--labels sample1 sample2 sample3 sample4 sample5 sample6 sample7 sample8 sample9 --read_group file_name barcode --sqanti_output --threads 24 --mode tenX_v3
--barcoded_reads sample1_reads_CB_UMI.tsv sample2_reads_CB_UMI.tsv sample3_reads_CB_UMI.tsv sample4_reads_CB_UMI.tsv sample5_reads_CB_UMI.tsv sample6_reads_CB_UMI.tsv sample7_reads_CB_UMI.tsv sample8_reads_CB_UMI.tsv sample9_reads_CB_UMI.tsv --model_construction_strategy default_pacbio --counts_format default

$Output

-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 643K Mar 12 01:25 islet_pacbio_n9.discovered_gene_counts.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 475M Mar 12 01:25 islet_pacbio_n9.discovered_gene_grouped_barcode_counts.linear.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 846M Mar 12 03:15 islet_pacbio_n9.discovered_gene_grouped_barcode_counts.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 1021M Mar 12 03:36 islet_pacbio_n9.discovered_gene_grouped_barcode_tpm.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 6.0M Mar 12 01:25 islet_pacbio_n9.discovered_gene_grouped_file_name_counts.linear.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 1.5M Mar 12 03:00 islet_pacbio_n9.discovered_gene_grouped_file_name_counts.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 4.2M Mar 12 03:00 islet_pacbio_n9.discovered_gene_grouped_file_name_tpm.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 719K Mar 12 03:00 islet_pacbio_n9.discovered_gene_tpm.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 2.2M Mar 12 01:24 islet_pacbio_n9.discovered_transcript_counts.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 548M Mar 12 01:25 islet_pacbio_n9.discovered_transcript_grouped_barcode_counts.linear.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 2.5G Mar 12 02:08 islet_pacbio_n9.discovered_transcript_grouped_barcode_counts.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 2.7G Mar 12 03:00 islet_pacbio_n9.discovered_transcript_grouped_barcode_tpm.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 17M Mar 12 01:25 islet_pacbio_n9.discovered_transcript_grouped_file_name_counts.linear.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 4.3M Mar 12 01:25 islet_pacbio_n9.discovered_transcript_grouped_file_name_counts.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 12M Mar 12 01:25 islet_pacbio_n9.discovered_transcript_grouped_file_name_tpm.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 2.4M Mar 12 01:25 islet_pacbio_n9.discovered_transcript_tpm.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 1.1G Mar 12 03:36 islet_pacbio_n9.extended_annotation.gtf
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 796K Mar 11 23:14 islet_pacbio_n9.gene_counts.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 703M Mar 11 23:14 islet_pacbio_n9.gene_grouped_barcode_counts.linear.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 917M Mar 11 23:31 islet_pacbio_n9.gene_grouped_barcode_counts.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 1.2G Mar 11 23:58 islet_pacbio_n9.gene_grouped_barcode_tpm.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 6.1M Mar 11 23:14 islet_pacbio_n9.gene_grouped_file_name_counts.linear.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 1.6M Mar 11 23:14 islet_pacbio_n9.gene_grouped_file_name_counts.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 4.4M Mar 11 23:14 islet_pacbio_n9.gene_grouped_file_name_tpm.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 906K Mar 11 23:14 islet_pacbio_n9.gene_tpm.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 14M Mar 12 03:36 islet_pacbio_n9.novel_vs_known.SQANTI-like.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 961M Mar 12 03:36 islet_pacbio_n9.read_assignments.tsv.gz
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 4.1M Mar 11 23:14 islet_pacbio_n9.transcript_counts.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 400M Mar 11 23:14 islet_pacbio_n9.transcript_grouped_barcode_counts.linear.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 2.4G Mar 12 00:38 islet_pacbio_n9.transcript_grouped_barcode_counts.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 2.5G Mar 12 01:24 islet_pacbio_n9.transcript_grouped_barcode_tpm.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 12M Mar 11 23:14 islet_pacbio_n9.transcript_grouped_file_name_counts.linear.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 3.6M Mar 11 23:14 islet_pacbio_n9.transcript_grouped_file_name_counts.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 9.3M Mar 11 23:15 islet_pacbio_n9.transcript_grouped_file_name_tpm.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 244M Mar 12 03:36 islet_pacbio_n9.transcript_models.gtf
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 4.8M Mar 11 23:14 islet_pacbio_n9.transcript_tpm.tsv
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 950M Mar 11 21:53 islet_pacbio_n9.UMI_filtered.ED3.allinfo.gz
-rw-r--r--. 1 suhailakhtar.ansari OG-IDC-User 407 Mar 11 21:53 islet_pacbio_n9.UMI_filtered.ED3.stats.tsv

[andrewprzh]

Dear @Suhail221107

Could you also send the log file?

The per barcode counts should be in *barcode_counts.tsv. If the number of barcodes is sufficient (>100), they are automatically converted to mtx format. It also might be a problem in 2.2.1, I remember fixing something for the next version 2.3, which will be released today.

In any case, you can covert *_barcode_counts.linear.tsv to mtx using src/convert_grouped_counts.py --output_format mtx script.

Best
Andrey

[Suhail221107]

Hi Andrey,

Thank you so much for the quick reply and the detailed explanation.

I have attached the log file to this email.
I do see the *_barcode_counts.tsv file in my output folder, but it was not converted to mtx, so I will try the convert_grouped_counts.py --output_format mtx script on *_barcode_counts.linear.tsv as you suggested.

It's great to hear that the fix will be included in version 2.3, released today. I will also give the new version a try once it is available.
​

Best regards,
Suhail

isoquant.log

[andrewprzh]

The log seems to be alright, so it was the conversion bug. IsoQuant was converting counts to usual TSV matrix instead of MTX.

The new release is out, but I you can still use the conversion script to save time.

Best
Andrey

[Suhail221107]

Hi Andrey,

Thanks for the quick fix and for looking into the logs. Good to know it was the conversion bug and not something on my end.
I'll go ahead and use the conversion script for now and switch to the new release when I get a chance.

Best,
Suhail

[Suhail221107]

Hi Andrey,

I tested the new version (2.3) with our dataset, and the run completed successfully. I have a question regarding the post-processing of the output files. I understand that IsoQuant already produces SQANTI3-like output, and I have generated those as well. However, following our collaborator’s suggestion, I’d like to run the SQANTI3 QC/Filter step again on the IsoQuant output.

Would it be possible to run the SQANTI3 filtering directly on the IsoQuant results? Also, how can we map the filtered transcripts back to our cell barcodes, UMI IDs, or read IDs?

Best,
Suhail

[andrewprzh]

Dear @Suhail221107

Transcripts are discovered by using data as bulk. To trace it back to barcodes or reads, you can:

1. Use --read_group barcode to get per-barcode counts for each transcript (default behavior in the next release);
2. Add --large_output read2transcripts to generate TSV with read ids and corresponding transcripts;
3. Run IsoQuant again providing new GTF as a reference annotation and use read_assignments.tsv.gz (probably a bit more accurate, than 2).

Overall, I suggest discovering novel transcript using bulk more and then running single-cell mode using generated GTF with --no_model_construction. If you discover novel isoforms in single-cell/spatial mode, it will filter out reads corresponding to novel/unknown genes during UMI deduplication.

Best
Andrey