[maker-devel] keep_preds values

[Mikael Brandström Durling]

Ok. But that is not implemented no as far as I can tell from the source, right? Or is it reflected in the AED for the unsupported models?

Mikael

10 mar 2014 kl. 16:59 skrev Carson Holt <carsonhh at gmail.com>:

> Yes.  It will eventually perform an AED like calculation between multiple
> predictors (i.e. if you use 3 predictors it, then you require support by
> at least 2 predictors across all exons to get a value of 0.33).  A value
> of 0 would be perfect concordance across all 3 predictors.
> 
> —Carson
> 
> 
> 
> 
> On 3/10/14, 7:57 AM, "Mikael Brandström Durling" <mikael.durling at slu.se>
> wrote:
> 
>> Hi Carson and Daniel,
>> 
>> That sounds more logical to me.  Then it would be appropriate to change
>> the comment of keep_preds in the generated config files.
>> 
>> Would it make sense to make keep_preds a non-binary value to evaluate the
>> concordance between ab initio models obtained from different predictors?
>> That would assume that it is less likely to be a false positive when two
>> or more predictors suggest the same unsported model?
>> 
>> Mikael
>> 
>> 
>> 10 mar 2014 kl. 16:51 skrev Carson Holt <carsonhh at gmail.com>:
>> 
>>> Actually that is false. The keep_preds option is still binary.  Any
>>> value
>>> other than 0 sets it to true.  There was discussion about making it a
>>> non-binary value, but that has not been implemented.
>>> 
>>> —Carson
>>> 
>>> 
>>> On 3/10/14, 7:47 AM, "Daniel Ence" <dence at genetics.utah.edu> wrote:
>>> 
>>>> Hi Mikael, 
>>>> 
>>>> The keep_preds parameter is often used the same as a binary parameter,
>>>> but it doesn't have to be. The concordance that is mentioned in the
>>>> comment line is the AED for that prediction. AED is a measurement of
>>>> how
>>>> well a prediction is supported by the evidence and ranges from 0 - 1. A
>>>> prediction with an AED of 0 matches the evidence exactly while a
>>>> prediction with an AED of 1 isn't overlapped by any evidence.
>>>> 
>>>> The default behavior for MAKER is to make a gene model out of a
>>>> prediction with any AED <1. When you change the keep_preds option from
>>>> 0
>>>> to 1, then MAKER will make a gene model out of any prediction that
>>>> matches the other parameters (like single_exon, min_exon, etc). Setting
>>>> the keep_preds option to somewhere in between 0 and 1 will set a
>>>> ceiling
>>>> on the AED required for promoting a prediction to a gene model.
>>>> 
>>>> From a user standpoint, when you will almost certainly lose gene models
>>>> when you set AED at an intermediate value, but you might benefit by
>>>> knowing that all your models will now have an AED of at least a certain
>>>> value. 
>>>> 
>>>> I hope that helps; let me know if it didn't.
>>>> 
>>>> ~Daniel
>>>> 
>>>> PS The original paper that described the AED is Eilbeck et al in BMC
>>>> Bioinformatics 2009. It's also discussed in more detail in the MAKER2
>>>> paper, the MAKER-P paper, and the Yandell and Ence Nature Reviews
>>>> Genetics paper from 2012.
>>>> 
>>>> Daniel Ence
>>>> Graduate Student
>>>> Eccles Institute of Human Genetics
>>>> University of Utah
>>>> 15 North 2030 East, Room 2100
>>>> Salt Lake City, UT 84112-5330
>>>> ________________________________________
>>>> From: maker-devel [maker-devel-bounces at yandell-lab.org] on behalf of
>>>> Mikael Brandström Durling [mikael.durling at slu.se]
>>>> Sent: Monday, March 10, 2014 4:27 AM
>>>> To: maker-devel at yandell-lab.org
>>>> Subject: [maker-devel] keep_preds values
>>>> 
>>>> Hi,
>>>> 
>>>> Can someone, please, explain the keep_preds parameter, as it works now
>>>> with a value between 1 and 0? It used to be binary, but now it seems to
>>>> test concordance towards something. The maker wiki doesn’t explain it
>>>> any
>>>> further either.
>>>> 
>>>> Thanks,
>>>> Mikael
>>>> 
>>>> 
>>>> _______________________________________________
>>>> maker-devel mailing list
>>>> maker-devel at box290.bluehost.com
>>>> http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org
>>>> 
>>>> _______________________________________________
>>>> maker-devel mailing list
>>>> maker-devel at box290.bluehost.com
>>>> http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org
>>> 
>>> 
>> 
> 
>