input for sqanti3 rescue --refClassif

[mhoang22]

Hi SQANTI3 team,
Im running SQANT3 version 5.5.3, and interested in running mode full for the rescue step. I have read the wiki , and have a couple of questions about the --refClassif flag

(The help command reads:
-k REFCLASSIF, --refClassif REFCLASSIF Full path to the classification file obtained when running SQANTI3 QC on the reference transcriptome. Mandatory when running the rescue on full mode
)

Set up:
Previously, I have run SQ3 QC using reference fasta using the command below, then SQ3 filter (ML) .
sqanti3_qc.py --isoforms ${ISOFORMS} \ --refGTF ${REF_GTF} \ --refFasta ${REF_GENOME_FA} \ --CAGE_peak ${CAGE} \ --polyA_motif_list ${POLY_A} \ -o ${OUTPUT_PREFIX} -d ${OUTPUT_DIR} -fl ${FL_COUNT} \ --short_reads T3_short_reads.fofn --cpus ${CPU_NUM} --report both

Questions:

To run SQ3 rescue mode full, do I need to rerun the SQ3 QC step with some different inputs? - if it is the case, what are the inputs I have to swap out (GTF? REF_GENOME_FA ? )?

Or am I supposed to just provide the path to the SQ3 QC output from the command above?

Thanks in advance for your help!
My

[pabloati]

Hi My,

That parameter referes to the classification file that SQANTI3 produces when running the reference against itself. You do not need to run SQ3 filter or anything. It is important that you use the same orthogonal data (CAGE peaks, poly A motifs) as you did on your original run. In your case above, it would be like this:

sqanti3_qc.py --isoforms ${REF_GTF} \ --refGTF ${REF_GTF} \ --refFasta ${REF_GENOME_FA} \ --CAGE_peak ${CAGE} \ --polyA_motif_list ${POLY_A} \ -o ${OUTPUT_PREFIX} -d ${OUTPUT_DIR} -fl ${FL_COUNT} \ --short_reads T3_short_reads.fofn --cpus ${CPU_NUM} --report both

Let me know if that makes more sense. :)

[mhoang22]

Thanks so much for your reply!

I took a look at the command you sent above. So my understanding is that I will need to 'run the reference against itself' by supplying the ${REF_GTF} for the --isoforms (instead of the long read collapsed gff originally), and the rest are the same, correct?

Would supplying the short read data still be critical? / Would kallisto and STAR be run internally to map the short read data to the ${REF_GTF} if i supply the .fofn ? (Im hoping yes, so sqanti could be aware of the transcript models which might have been missed in the long read because of low coverage, but abundant in short read data)

[mhoang22]

Also I think that --fl ${FL_COUNT} can be excluded from the command since the --isoforms now have the ${REF_GTF}, not the long read (pacbio/isoseq) gff anymore?

[pabloati]

Hi My,

Yes, you should hange the --isoforms from the long read collapsed gtf to the reference GTF. Keep everything the same, even the short reads (that way you can also see how well the short reads from your experiment match the reference), as the same filter is applied. However, you are completely right that the --fl option can be excluded! :)

[mhoang22]

just want to update that i followed your recommended command (minus the fl flag), everything ran smoothly, and I was able to generate the output for rescue step as needed. thanks so much for your help!