augustus gene prediction

Question

[Shakunthala Natarajan]

I have a genome of a novel fungus. I need to predict the genes in it using Augustus. In the --species flag of augustus, what should I specify, since my fungus is a novel one and does not match any of the identifiers provided in the readme doc in augustus github page.

Can someone please help me with this? Thank you!

Answer 1

[liorglic]

If none of the available profiles is close enough to your species, then you might consider training Augustus yourself. While this might be a little difficult, I find that using BUSCO v4 (I think newer versions don't use augustus anymore) can make it much easier. BUSCO will automatically train Augustus and make a new profile which you can then use to predict genes. Another thing you should note is that ab-initio gene prediction (which is what Augustus does) is a noisy and often inaccurate process. It is usually only one step in a large genome annotation pipeline. To get a good annotation, you'd probably want to use evidence such as transcripts, RNA-seq, and protein sequences, in the context of an annotation pipeline like MAKER, PASSA, or EvidenceModeler.

Answer 2

[Mensur Dlakic]

There are many fungal species that are known to augustus:

https://bioinf.uni-greifswald.de/augustus/

Usually you select the one known or suspected to be closest to your species of interest. It is not a bad idea to try more than one.