A small question about collapse isoforms script about cdna_cupcake

[Magdoll]

Hi @WayneForever ,

Yes you are correct that the estimation of genes (technically, just "loci") changes.

This is because the collapse script doesn't actually know where genes are since no annotation is used. It simply creates "loci", where the definition is "a collection of overlapping transcripts on the same strand, even if they only overlap by 1 bp"

So, with more transcripts, it's possible now there are more overlaps and hence fewer "genes".

--Liz

[WayneForever]

Thanks for your reply, Liz.

[WayneForever]

Hi, @Magdoll

Sorry for bothering you again. There is another thing that I don't understand.

This project follows the Isoseq2 from Pacbio where 16 cells were sequenced. In total, I got 1,267,862 reads and 288,773 isoforms based on 'collapse_isoforms_by_sam.py' script. The resulting file 'all.collapsed.group.txt' shows that some loci (The gene I mentioned above) has even hundreds of isoforms such as PB.1.259 or PB.63.242.

In most circumstances, this situation seldom happens because the number of isoforms in one gene shouldn't be that much. Can this situation happen if my input of the script is correct?

My script runs as follows, and the file 'rm_fusion.mapped.sorted.sam' is the gmap.sam without fusion gene:
python collapse_isoforms_by_sam.py --input NGS_corrected.fasta -s rm_fusion.mapped.sorted.sam -o all

Sorry again for my bothering and thank you very much.

[Magdoll]

Hi @WayneForever ,

So you had 288,773 HQ isoforms (after running Iso-Seq2 pipeline) and then ran collapsed?

I would look at the before (rm_fusion.mapped.sorted.sam) and after (all.collapsed.gff) on a browser (ex: IGV) and look if the high splicing diversity for those genes is really there. It's entirely possible, especially for neural genes in human, to have a lot of isoforms. One of the first Iso-Seq papers found 247 neurexin 1 isoforms: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3977267/

--Liz

[WayneForever]

The 288,773 isoforms are the results of collapse isoform. My input is the quivered fasta after Iso-Seq2 pipeline. The species of the project is Macaca.
Now I'm thinking that this result may be caused by high redundancy in the quivered fasta. And I'm try to reduce redundancy now. Thanks anyway. :)

[Magdoll]

Good to hear. Let me know what you find out.