Short exons account for a large proportion

[linyuiz]

Thank you for developing Helixer.

I tested several plants and found that a significant proportion of them have very short exons.
After filtering these genes, I directly reduced the number of genes from 58,000 to 48,000.
Is this a bug for Helixer?
Are there any improved methods in the newer versions of Helixer?

[felicitas215]

Hello,
thank you for your comment and the extensive research and nice graphs you provided.

Your issue relates to an existing thread/issue #138. I would recommend checking this one out, but to answer your question here in short as well:
This is a known issue of Helixer or more accurately HelixerPost trying to fit mistakes Helixer makes in the phase predictions due to possibly INDELS into its strict phase counting by introducing 'fake' 1 or 2 base exons to compensate for the missing / extra base in the genome. Another case we have seen is the UTR not being predicted confidently by Helixer, which leads to a change from CDS or intron directly to the intergenic region without a UTR in between. HelixerPost, adhering to a strict pattern, inserts a 1 base UTR/exon to compensate for this. We are working on improving this behaviour, but there are currently no new versions of Helixer that improve on this.