What'S Your Pipeline For Gene Prediction On Prokaryotic Contigs ?

Question

[Pasta]

Hi,

There a few tools available to predict genes on bacterial genomes such as Glimmer, Prodigal and GeneMarkS. I tried them but they always make divergent predictions, so I was wondering if you are using some kind of pipeline to make a consensus gene prediction? More generally, what do you use for this gene prediction on prokaryotes ?

Answer 1

[Lee Katz]

I am going to give another biased answer, CG-Pipeline, which is available at nbase.biology.gatech.edu. I am currently the lead developer for CG-Pipeline. It uses a combination of GeneMark, Glimmer3, and BLAST vs SwissProt and outputs a GenBank file with gene predictions for CDS. Additionally, it uses RNAmmer and tRNAscan-SE to detect RNA genes (these genes also appear in the output GB file). Our publication discusses the algorithm we use in depth, but basically a gene is called if two out of three predictors call it, and the longest prediction is used. CG-Pipeline has been used for several bacterial pathogens including but not limited to Neisseria, Haemophilus, Bordetella, E. coli, and Vibrio.

CG-Pipeline has modules for assembly, gene prediction, and annotation, but you can readily use only the run_prediction script if you want. In other words, it is very modular, and you can pick and choose the scripts you want to use, or you can run the entire thing in one command.

Answer 2

[Zev.Kronenberg]

Our lab developed MAKER and that is what we use. It combines multiple gene prediction algorithms and evidence to create the "best fitting gene model." MAKER is widely used. Of course I am biased.

[Michael]

Isn't maker for eukaryotes only? DIYA is said to be for prokaryotes, according to the gmod web-site, or have you made a new version of MAKER for both?