[BioC] plotpfm has been removed in PWMEnrich?

[Robert Stojnic]

Hi Fabrice,

The negative values are usual as the PWM matrix is defined as log2 
probability_base / probability_background. So if the any of the 
probabilities is smaller than 0.25 in your input matrix, this will give 
a negative value. Maybe you wanted to get the heights of the individual 
letters in the logo?

# convert the count matrix to probabilities
p = t(t(pfm) / colSums(pfm))

# this is without the small-sample correction, for more information see 
https://en.wikipedia.org/wiki/Sequence_logo or the original paper
h = - colSums(p * log2(p))
heights = t(t(p) * (2-h))

All the transpositions (calls to t()) are needed to make sure the matrix 
and vector operations are carried out along the correct dimensions.

Cheers, Robert

On 10/12/13 16:41, Fabrice Tourre wrote:
> PWM always give me some negative value. This strange to me.
>
> PFMtoPWM(pfm)
> An object of class 'PWM'
> ID:
> Target name:
> Frequency matrix:
> $pfm
>       [,1]    [,2]    [,3]    [,4]    [,5]    [,6]    [,7]    [,8]    [,9]
> A 3289112 3286601 3324808 3295086 3297953 3371158 3350851 3380593 3406893
> C 2770671 2771901 2749093 2791942 2791869 2756225 2779305 2780189 2727296
> G 2836085 2853221 2826205 2830413 2845211 2809652 2778112 2742323 2542202
> T 3286234 3270379 3281996 3264661 3247069 3245067 3273834 3278997 3505711
>      [,10]   [,11]   [,12]   [,13]   [,14]   [,15]   [,16]   [,17]   [,18]
> A 3293358 2178674 4352405 3227324 3461232 3446487 3259436 3276060 3271729
> C 2663331 2932504 2139363 3052255 2674074 2550187 3005584 2805702 2844711
> G 2467603 2413635 1644787 2028507 2906055 2827872 2948758 3050624 2971989
> T 3757810 4657289 4045547 3874016 3140741 3357556 2968324 3049716 3093673
>      [,19]   [,20]   [,21]
> A 3297076 3293868 3334010
> C 2853311 2824444 2768038
> G 2891960 3017416 2974502
> T 3139755 3046374 3105552
> Position weight matrix (PWM):
> $pwm
>          [,1]        [,2]       [,3]       [,4]        [,5]        [,6]
> A  0.1110069  0.10990513  0.1265798  0.1136249  0.11487965  0.14655305
> C -0.1364558 -0.13581544 -0.1477355 -0.1254222 -0.12545992 -0.14399751
> G -0.1027904 -0.09409968 -0.1078251 -0.1056786 -0.09815553 -0.11629972
> T  0.1097440  0.10276665  0.1078823  0.1002420  0.09244685  0.09155707
>          [,7]       [,8]       [,9]      [,10]      [,11]      [,12]
> A  0.1378363  0.1505851  0.1617654  0.1128682 -0.4832408  0.5151216
> C -0.1319670 -0.1315082 -0.1592199 -0.1934594 -0.0545581 -0.5095098
> G -0.1325864 -0.1512927 -0.2606125 -0.3035809 -0.3354836 -0.8887903
> T  0.1042900  0.1065634  0.2030159  0.3032009  0.6127992  0.4096436
>           [,13]       [,14]      [,15]       [,16]        [,17]       [,18]
> A  0.083647231  0.18459445  0.1784354  0.09793116  0.105270587  0.10336206
> C  0.003184313 -0.18765178 -0.2560881 -0.01904584 -0.118329389 -0.09840908
> G -0.586272844 -0.06762917 -0.1069744 -0.04658375  0.002413189 -0.03526240
> T  0.347138703  0.04441379  0.1407203 -0.03704261  0.001983716  0.02262953
>          [,19]         [,20]       [,21]
> A  0.11449595  0.1130915480  0.13056724
> C -0.09405417 -0.1087242780 -0.13782742
> G -0.07464358 -0.0133775737 -0.03404303
> T  0.04396080  0.0004018867  0.02815854
> With background nucleotide frequencies which also serve as pseudo-count:
> $prior.params
>     A    C    G    T
> 0.25 0.25 0.25 0.25
>
> On Tue, Dec 10, 2013 at 7:29 AM, Robert Stojnic <rainmansr at gmail.com> wrote:
>> Dear Fabrice,
>>
>> You can plot the PFM in PWMEnrich by converting it to PWM:
>>
>> plot(PFMtoPWM(pfm_matrix))
>>
>> Unfortunately the old function was deprecated. Sorry if it broke your code!
>>
>> Cheers, Robert
>>
>>
>> On 07/12/13 04:41, Fabrice Tourre wrote:
>>> Dear expert,
>>>
>>> I just want to plot PFM using PWMEnrich package. It seems seqlogo
>>> cannot do this. But I found in PWMEnrich2.6.2, plotpfm does not exist.
>>> How can I plotPFM using new PWMEnrich?
>>>
>>> _______________________________________________
>>> Bioconductor mailing list
>>> Bioconductor at r-project.org
>>> https://stat.ethz.ch/mailman/listinfo/bioconductor
>>> Search the archives:
>>> http://news.gmane.org/gmane.science.biology.informatics.conductor
>>>