Strand issues for several novel genes in GTF (".", not "+" or "-")

[jjuhyunkim]

Hi,

I encountered an error stating assert strand == '+' or strand == '-' while running sqanti3_qc.py. Additionally, I noticed several novel genes with a strand designation of "." in the 7th column in transcript_models.gtf (figure below)

These genes are not mono-exonic.
I utilized the most recent version of isoquant (v3.3.1)

I have notice that similar issues in the past, and you addressed them in a previous version.
However, the problem persists.
Could you assist me in identifying the underlying cause of this issue?

Thank you! : )
JH

[andrewprzh]

Dear @Juhyun-Kim-0203

I suspect that these transcripts may not have any canonical splice sites, and thus their strand cannot be determined.
Could you tell which genome you are using? Also, I would be helpful if you could share a small subset of reads that map these transcripts. Is it possbile?

Probably, I should also make an option for filtering out such "unreliable" transcript predictions.

Best
Andrey

[andrewprzh]

Dear @Juhyun-Kim-0203

Found a silly bug, it happens when lower-case characters appear in the genome (masked regions). Fixed now in master and will be fixed in the next release, which I plan to do ASAP.

Thanks for the report!

Best
Andrey

[jjuhyunkim]

Dear Andrey Thank you for your rapid response! Is it okay with hard masking? Thank you! Juhyun

…

Dear @Juhyun-Kim-0203 <https://github.com/Juhyun-Kim-0203> Found a silly bug, it happens when lower-case characters appear in the genome (masked regions). Fixed now in master and will be fixed in the next release, which I plan to do ASAP. Thanks for the report! Best Andrey — Reply to this email directly, view it on GitHub <#107 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A7I5WUG4FAI3JYYF6AQRXZLXYCYTNANCNFSM6AAAAAA32MONIU> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[andrewprzh]

Dear @Juhyun-Kim-0203

For hard-masked genomes I presume minimap won't be able to map the reads and thus the transcript won't be discovered at all. However, if splice sites have "NN" characters the strand will be also reported as ".".

Best
Andrey

[andrewprzh]

Finally released new version 3.4, which fixes this issue.