questions about flair align and transriptome

[a927050047]

I am using flair align and transriptome to generate gff files. During the comparison, I encountered some unexpected results with the Flair pipeline and would appreciate your insights.
commands:
FLAIR:

flair align -t 20 -g diploid.f.fa \
    -r ../filt_pychopper/$sample.pass.fq -t 20 \
    --filtertype separate --remove_singleexon -o $sample.align

flair transcriptome -t 20 -g diploid.f.fa \
    --junction_tab LWR11_SHORTREAD.SJ.out.tab \
    -b $sample.align.bam  -o $sample.transc \
    -w 50 -s 1 --filter ginormous   -n longest --max_ends 4 --keep_sup --keep_intermediate

minimap2 + StringTie:

minimap2 -t 20 -ax splice:hq -u f  LWR11.diploid.f.fa filt_pychopper/$sample.pass.fq \
        | samtools view -bS - | samtools sort -o bam/$sample.bam
stringtie -L -o stringtie/$sample.gtf --conservative -m 50  -A stringtie/$sample.txt -p 12 bam/$sample.bam 

visualization using IGV:

Observations and Questions:

1. Incomplete Isoform Reconstruction by Flair:​ Based on IGV visualization, Flair failed to reconstruct a full-length isoform for a particular gene, even though there appears to be sufficient read support. The Minimap2+StringTie pipeline, in contrast, generated a complete isoform. Could you suggest potential reasons why Flair might not have assembled the full-length transcript in this case?
2. Exon Misplacement by flair align:​ The flair alignstep seems to have misplaced a 16bp exon. Is there a recommended way to correct such alignment errors, perhaps through parameter adjustment or a specific filtering strategy?

[diekhans]

Thanks for the detailed report. Would you be able to share the input to FLAIR so we can reproduce the problem? a927050047 ***@***.***> writes:

…

I am using flair align and transriptome to generate gff files. During the comparison, I encountered some unexpected results with the Flair pipeline and would appreciate your insights. commands: FLAIR: ``` flair align -t 20 -g diploid.f.fa \ -r ../filt_pychopper/$sample.pass.fq -t 20 \ --filtertype separate --remove_singleexon -o $sample.align flair transcriptome -t 20 -g diploid.f.fa \ --junction_tab LWR11_SHORTREAD.SJ.out.tab \ -b $sample.align.bam -o $sample.transc \ -w 50 -s 1 --filter ginormous -n longest --max_ends 4 --keep_sup --keep_intermediate ``` minimap2 + StringTie: ``` minimap2 -t 20 -ax splice:hq -u f LWR11.diploid.f.fa filt_pychopper/$sample.pass.fq \ | samtools view -bS - | samtools sort -o bam/$sample.bam stringtie -L -o stringtie/$sample.gtf --conservative -m 50 -A stringtie/$sample.txt -p 12 bam/$sample.bam ``` visualization using IGV: <img width="1865" height="765" alt="image" src="https://github.com/user-attachments/assets/6f5aa85c-623b-404d-87bc-5163c6369a61" /> <img width="1672" height="745" alt="image" src="https://github.com/user-attachments/assets/2b4db123-7cab-4a54-a622-719e0010554b" /> Observations and Questions: 1. Incomplete Isoform Reconstruction by Flair:​ Based on IGV visualization, Flair failed to reconstruct a full-length isoform for a particular gene, even though there appears to be sufficient read support. The Minimap2+StringTie pipeline, in contrast, generated a complete isoform. Could you suggest potential reasons why Flair might not have assembled the full-length transcript in this case? 2. Exon Misplacement by flair align:​ The flair alignstep seems to have misplaced a 16bp exon. Is there a recommended way to correct such alignment errors, perhaps through parameter adjustment or a specific filtering strategy? -- Reply to this email directly or view it on GitHub: #617 You are receiving this because you are subscribed to this thread. Message ID: ***@***.***>

[a927050047]

Sorry for the delay, I extracted this loci and re-run falir, the transcriptome result is kinda different, but I hope this could help. @diekhans
region.zip

[diekhans]

If you could extract the region from original flair align, BAM that would be useful too. Due to holidays, we might be a bit delayed getting back to you a927050047 ***@***.***> writes:

…

Sorry for the delay, I extracted this loci and re-run falir, the transcriptome result is kinda different, but I hope this could help. @diekhans [region.zip](https://github.com/user-attachments/files/23920207/region.zip) <img width="1117" height="365" alt="image" src="https://github.com/user-attachments/assets/64dcfb4e-122a-4058-a546-3ba6212ffeeb" /> -- Reply to this email directly or view it on GitHub: #617 (comment) You are receiving this because you were mentioned. Message ID: ***@***.***>

[a927050047]

These are bam and gff extracted from flair align and transcriptome.
origin.zip