Genome annotation with BRAKER: how to interpret the results?

[svitlana]

Hello,

I am new to the genome annotation and I'm lost with the interpretation of results produced by BRAKER.

• I have a de novo assembly of an insect genome (N50 = 350kb, length = 1.9 Gb).
• I masked the repeats using RepeatModeler and RepeatMasker.
• I mapped the RNA-Seq data from the same species to the (hard) masked genome with HISAT2.
• I used BRAKER to annotate my (soft masked) genome with the bam file produced by HISAT2.

I have 54000 entries in the resulting augustus.hints.gff file. That means that Augustus predicted 54k genes, right? We expect to have between 10k and 20k genes for our species, so I would like to understand why there are so many genes in our prediction.

Among these 54k entries, 38k entries contain the following information:

# % of transcript supported by hints (any source): 0

Does it mean that these predictions are of poor quality and I should only keep predictions with a significant %?

Any other suggestions on how to enhance the annotation of my genome are welcome!