Training Augustus for Maker

Question

[Rox]

Hi everyone !

I'm aiming to fully assemble and annotate the D. suzukii genome. For the annotation part, it is my first time, I already asked a lot of questions here, and get a lot of answers, but some points remains very unclear to me, even in the Maker tutorial (http://weatherby.genetics.utah.edu/MAKER/wiki/index.php/MAKER_Tutorial#Merge.2FResolve_Legacy_Annotations ).

In their tutorial, they fully described the procedure to train SNAP. So we have to launch several times maker in order to produce a final pyu.hmm, corresponding to the "final" SNAP model that I'm going to give to my "final" maker.

I was looking for something similar for Augustus, but couldn't really find what I wanted.

So what I did, is to run myself Augustus, first abinitio, and then giving hints (some RNAseq data that correspond to my organism).

I was aiming to give theses 2 augustus output file to maker using for example the maker_opts.ctl :

 ...
est_gff= #aligned ESTs or mRNA-seq from an external GFF3 file
...
pred_gff= #ab-initio predictions from an external GFF3 file
model_gff= #annotated gene models from an external GFF3 file (annotation pass-through)
...
other_gff= #extra features to pass-through to final MAKER generated GFF3 file

But as you see, there is a lot of places where I could provide my augustus GFF, and I'm not really confident about how theses parameters are similar, or how they affect the final annotation... Also not sure that I'm doing the right thing, and that I'm really "training" augustus or not.

Can I have some advices please ?

Cheers,

Roxane

[alslonik]

Hi, Roxane!

I am annotating my plant genome now and am working with Maker and the same detailed tutorial. My question is re: SNAP now. How do you actually know that it is enough to train and you can run your final Maker run. I have tried to run it several time and there is a difference in the number of genes every time. It is actually a kind of sinusoid graph - number of genes are going up and down... So when do you stop? Or how do you know that SNAP is trained? Do you wait until the plateau? How many times did you do the training and why?

Answer 1

[Juke34]

Hello ! Me again ...

You cannot train Augustus within MAKER. You can only create an annotation with MAKER that will be used to train Augustus aside.

Here links that explain how train Augustus:
gene set filter/selection for training ab initio annotation tools
http://avrilomics.blogspot.se/2013/04/training-augustus-gene-finding-software.html http://www.vcru.wisc.edu/simonlab/bioinformatics/programs/augustus/docs/tutorial2015/training.html

But I advise you to not train it ! Yes, I'm quite sure that the Drosophila melanogaster hmm profile already available into Augustus will be better that the one you will create.

[Rox]

Sorry for the late reply, I was busy with an other project, I'm currently reading the links you advised me, and they helped me a lot. Thanks !

Answer 2

[mafireyi]

Hi. You can use BRAKER to train Augustus. See a previous discussion on this here

A: Braker output into maker (Annotation)

[Juke34]

As they says in the paper, If you have only EST, BRAKER is better, but If you have EST and proteins, using MAKER still gives a better result.

[mafireyi]

and if u have proteins from a closely related organism?

[Juke34]

Yes ! Most of times it is like that, you don't have the proteins of the species you want to annotate.

Answer 3

[Juke34]

You can also use BUSCO to train Augustus

[kcamnairb]

I have run BUSCO with the "--long" option and it seemed to have completed successfully, but I don't see where the optimized augustus metaparameters are stored. I have looked in the augustus config directory and the BUSCO output directory. Is there somewhere else it is placed?

[Juke34]

It should be in $AUGUSTUS_CONFIG_PATH/species/yourspecies folder I guess.

[bifobear]

Did you ever figure this out? I'm trying to train Augustus with BUSCO as well, and despite augustus.log file claiming "Writing exon model parameters [1] to file ..." in AUGUSTUS_CONFIG_PATH, there is no output that I can find.

[Panos]

I know it's been a really long time since this was posted but I had to deal with this myself a couple of days ago.

So, the Augustus training parameters generated by BUSCO when you use the --long option, are saved in the current working directory (i.e. the directory from which you called BUSCO); they are NOT saved in the Augustus config/species/ directory.

More specifically, it should be in the run_XXX/augustus_output/retraining_parameters/ directory, where XXX is the value you gave to the -o option when running run_BUSCO.py.

[Juke34]

It depends which version you use. Version 3 is like you said, version 1 is like I said, version 2 I don’t remember well but I think it was like version 1.

[Juke34]

The species folder where the hmm models are stored are located within the Augustus installation folder. You can try the command

which augustus

And you will know where to look at.