[Biopython] bio.motifs P-value on pssm searches

[Bartek Wilczynski]

Dear Marco,

the score you mention is in fact a log-odds score. it represents a
logarithm of the ratio between the probability of the sequence in question
being generated from the motif or from a random generator.

If you want to get some analog of a p-value (the probability of obtaining a
score of x or higher), you need to look into the score distributions in the
thresholds package. For example if you want to know what score corresponds
to a p-value of 0.05 for motif M you can do

thresholds.ScoreDistribution(M).threshold_fpr(0.05)

Please remember that the thresholds are computed approximately to a given
precision (in the scoreDistribution constructor).

Naturally, if you are searching in a sequence of length 1000, you should
expect ~20 cases, for this given fpr.

Hope that helps
Bartek


On Thu, Nov 14, 2013 at 1:30 PM, Marco Galardini
<marco.galardini at unifi.it>wrote:

> Dear biopythoners,
>
> the Bio.motifs search of PSSM is a really effective tool when dealing with
> regulatory motifs. When searching a pssm in a DNA sequence, a bit score is
> associated with each position; I was wondering if you have any gotchas to
> obtain a P- or E-value from such scores. I couldn't find any method in the
> package that does that but maybe I've missed something.
>
> Thanks for your help,
> Marco
>
> --
> -------------------------------------------------
> Marco Galardini, PhD
> Dipartimento di Biologia
> Via Madonna del Piano, 6 - 50019 Sesto Fiorentino (FI)
>
> e-mail: marco.galardini at unifi.it
> www: http://www.unifi.it/dblage/CMpro-v-p-51.html
> phone:  +39 055 4574737
> mobile: +39 340 2808041
> -------------------------------------------------
>
> _______________________________________________
> Biopython mailing list  -  Biopython at lists.open-bio.org
> http://lists.open-bio.org/mailman/listinfo/biopython
>
>


-- 
Bartek Wilczynski
==================
Institute of Informatics
University of Warsaw
http://www.mimuw.edu.pl/~bartek