Micro exons with single bases or two

[ke-shi]

Hi,

We used Helixer to predict genes in plant genome assemblies we have constructed. Approximately 30k genes were often predicted and complete BUSCO scores are always >95%. It seems prediction accuracy is very high. However, recently, we found ~30% genes predicted by Helixer had single-base exons or two-base. -- it may be micro exons. I am wondering such micro exons are frequently found in plant genomes. Or, do we need to customize parameters to eliminate them?

Here is an example with a single-base exon at 1519:

ch01      Helixer gene    1166    1953    .       +       .       ID=ch01g00002
ch01      Helixer mRNA    1166    1953    .       +       .       ID=ch01g00002.1;Parent=ch01g00002
ch01      Helixer exon    1166    1189    .       +       .       ID=ch01g00002.1.exon.1;Parent=ch01g00002.1
ch01      Helixer exon    1519    1519    .       +       .       ID=ch01g00002.1.exon.2;Parent=ch01g00002.1
ch01      Helixer exon    1707    1953    .       +       .       ID=ch01g00002.1.exon.3;Parent=ch01g00002.1
ch01      Helixer CDS     1187    1189    .       +       0       ID=ch01g00002.1.CDS.1;Parent=ch01g00002.1
ch01      Helixer CDS     1519    1519    .       +       0       ID=ch01g00002.1.CDS.2;Parent=ch01g00002.1
ch01      Helixer CDS     1707    1951    .       +       2       ID=ch01g00002.1.CDS.3;Parent=ch01g00002.1
ch01      Helixer five_prime_UTR  1166    1186    .       +       .       ID=ch01g00002.1.five_prime_UTR.1;Parent=ch01g00002.1
ch01      Helixer three_prime_UTR 1952    1953    .       +       .       ID=ch01g00002.1.three_prime_UTR.1;Parent=ch01g00002.1

Thanks,
Kenta

[alisandra]

Hi Kenta,

Thanks for the feedback. Microexons are not expected in plant genomes at the frequency Helixer predicts them and these are very likely predictive mistakes. It's a known challenge spot/issue for Helixer and occurs especially where there is a high level of uncertainty in the raw predictions. Unfortunately I don't have a quick parameter fix. Maybe increasing the --peak-threshold to improve the overall precision would help some. Hopefully we'll have have a better solution in the future.

[ke-shi]

Hi Alisandra,

Thank you for the reply. I look forward the solution!

Best,
Kenta

[Wan9299]

i had the same problem in plant genome annotation