Isoquant vs Flames (vs Kallisto-lr ?)

[reJELIN]

Ask away!

Hello Nanopore Community,

Why does wf single-cell pipeline is still using Flames as transcript quantification tool as it as been proven to detect less transcript than Isoquant or Bambu ?

As we can see below in this figure (from the LRGASP consortium) Bambu & Isoquant are far superior in term of genes quantification and discovery. Even though Flames is able to compute single-cell quantification, Bambu also propose a single-cell tool capable of doing it.

I see no real reason to keep using Flames apart the fact that it was the first to propose a single-cell quantification tool.

[nrhorner]

Hi @reJELIN

Thanks for your input. We are looking into alternative isoform assembly and quantification tools. I'll let you know when we have something ready.

[reJELIN]

It seems possible to run Isoquant from bam file. Do you think it is possible to decompose the tagged.bam in multiple bam for each barcodes and use Isoquant with each barcodes and parallelize by the number of cores available ? I'm kind of curious to see how different are the results from Flames ?

[nrhorner]

You can run the entire tagged BAM file through IsoQuant, as it provides per-read assignment output, allowing you to compare it with the wf-single-cell results. To clarify, we do not use FLAMES directly for isoform identification, but we have incorporated some of its code into our workflow.

[lianov]

@reJELIN : take a look into the nf-core/scnanoseq pipeline as a potential solution as well. Version 1.0.0 already has IsoQuant, and version 1.1.0 which is about to be released (available in dev) has included oarfish: https://github.com/nf-core/scnanoseq. Happy to chat more about it in the nf-core/scnanoseq Slack group.

[nrhorner]

Great thanks will take a look