[maker-devel] Fragmented annotation

[Ole Kristian Tørresen]

Hi all,
I have annotated a fish genome (about 700 Mbp total, 90 kbp N50 contig, 270 kbp N50 scaffold), where I get 96576 gene models, 67917 with default filtering (quality_filter.pl -d) and 67917 with standard filtering (quality_filter.pl -s). I chose to report all genes with AED less than 0.5  (27437) as the high quality set. 

However, I wonder a bit. One thing is that 70k genes cannot be correct for this species (it is not polyploid), and the correct number of genes should be a bit more than 20k I think. I suspect that many of my genes are fragmented, how can I fix this? I have tried searching the forum, but cannot find any good answers. Is there some parameters I can adjust?

I have used SwissProt/UniProt and a Trinity assembly of reads from several stages of embryo development  as evidence. I used SNAP with CEGMA, AUGUSTUS trained with BUSCO actinoptergyrii genes and GeneMark-ES in first pass, SNAP trained on first pass annotation and AUGUSTUS trained on the transcriptome and first pass annotation together with GeneMark for second pass annotation. 

Thank you.

Ole