Annotation Of Genes On New Genome

Question

[Gregor Rot]

Hi there,

i have a new genome and i am searching for genes. I use various tools for detecting genes from DNA (augustus, genemark) and i also have RNA-seq data that i map to the genome. To get isoforms i use tophat/cufflinks/cuffmerge. I also blast proteins to the genome from similar species and get hits.

It's difficult to understand how to merge all this information into a consensus to get good gene models, for example Genemark and Augustus models mostly overlap but also differ a lot of times.

Is there an interactive tool to manage and decide which models are more correct? I am thinking about something visual that would display overlapping information (gene models, protein blast hits) and interactively allow me to decide which models are "correct".

What tools do you use to integrate various information when annotating genes on a new genome? Can somebody point me to papers that describe pipelines for gene annotation?

Thanks, Gregor

Answer 1

[Michael]

And again recommending Maker, but: Maker is not a graphical tool, it is a pipeline of tools to create consensus gene calls. That is a good start for an automatic draft annotation imho, but not sufficient in the long run. Good genome annotation requires manual curation by dedicated experts, possibly with integration of data from experimentalists. To aid manual curation, set up a genome browser, e.g. Gbrowse2, Jbrowse, or best a complete model organism database. It will give you the advantage of storing your genome annotations in a consistent way. Having a web-based genome browser will enable biologists for collaboration. Collaboration is the key.

In the genome browser, create tracks for the different gene-predictions (Maker, Augustus, Genemark), a track for the alignments of ESTs, RNA-sequencing coverage, re-sequencing, conservations scores and maybe blast hits to other related organisms and what not. Also run tRNAScan-SE to predict tRNA, afaik Maker doesn't do this (does it Zev?). Now, work with this annotation for a while, together with the biologists, to see if they e.g. find the genes they are interested in in the annotation. They should provide suggestions for improving the existing annotations.

If you intend to initiate a real curation process of your genome, I recommend to visit something like the Biocuration conference to see how curators work. (Curation is a fulltime job it seems, a very important one, but possibly not the most attractive work for a bioinformatician). In the proceedings of Biocuration there were a few consensus gene callers; also approaches for community annotation are described there. For arriving at a high quality, reliable and stable annotation, I expect that it takes a few years, even with a bunch of experts working on the genome.

Counter argument: Hey, this is so old fashioned. This guy takes it way too seriously, just take the draft annotation and be happy...

[Zev.Kronenberg]

You are correct Michael, Maker does not run tRNAScan-SE.

Answer 2

[Gregor Rot]

Thanks for your answers, i also found a very nice overview of the methods and strategies for genome annotation in http://www.nature.com/nrg/journal/v13/n5/abs/nrg3174.html

Answer 3

[Zev.Kronenberg]

Yes, Maker can do this. You can manually curate the maker output (GFF3) in Apollo. Wrangle some friends into an annotation jamboree.

bias: My lab is responsible for Maker.

[Damian Kao]

You work in Sanchez's lab? Do you work on planarians too?

[Zev.Kronenberg]

Yandell lab. The planarian project was wrapping up before I joined the lab. Sanchez left right as I was joining the department.

Answer 4

[Lee Katz]

I believe that Maker does something like this, but I haven't used it in depth like this. http://gmod.org/wiki/MAKER