[maker-devel] Augustus training within Maker

[Barry Moore]

Dan is correct about the workings of the hint based communication between MAKER and the gene predictors - and just to clarify further.  MAKER never 're-trains' Augustus or any of the other gene predictors.  Many people will retrain the gene predictors between iterative MAKER runs, but this is a manual (and for Augustus non-trivial) step.

B

On Nov 15, 2012, at 4:34 AM, Eleanor Stanley wrote:

> Aha - that makes sense, so with every run Augustus is run with the BLAST hints available, thanks for clarifying this
> 
> Ele
> 
> 
> On 15 Nov 2012, at 11:14, Daniel Hughes wrote:
> 
>> heya,
>> 
>> Could you please confirm something for me:
>> 
>> If with the first Maker run you provide the location of the Augustus species config folder of gene parameters, then Augustus runs using this data alone.
>> I understand that with a second Maker run Augustus retrains using the BLAST evidence from the 1st run.
>> 
>> To achieve this, do I need to change anything in maker_opts.ctl or do I leave
>> augustus_species= #Augustus gene prediction species model
>> 
>> 
>> AFAIK maker runs augustus, blast etc., as sort of raw compute stages in the earlier tier_types, the results of these stages e.g. protein alignments etc., are then 'synthesised' into 'hints' (early tier_type 3) that are provided to ab initio predicters capable of taking external 'hints' - such as augustus - that are used to generate the final models (after further adjustmnets wrt., utrs etc.). that is to say i think what you mean by first and second maker runs are actually internal usage of ab initios that occur at different stages of any single maker run and you should be shielded from this.
>> 
>> dan.
>> 
>> 
>> 
> 
> 
> -- The Wellcome Trust Sanger Institute is operated by Genome Research Limited, a charity registered in England with number 1021457 and a company registered in England with number 2742969, whose registered office is 215 Euston Road, London, NW1 2BE. 
> _______________________________________________
> maker-devel mailing list
> maker-devel at box290.bluehost.com
> http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org

Barry Moore
Research Scientist
Dept. of Human Genetics
University of Utah
Salt Lake City, UT 84112
--------------------------------------------
(801) 585-3543




-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://yandell-lab.org/pipermail/maker-devel_yandell-lab.org/attachments/20121116/f572d106/attachment-0003.html>