Sum of MEME letter-probability matrix is larger than 1

[강병희]

Hi,

Thank your for the great database JASPAR. It is very useful for molecular studies.

I have used JASPAR to find DNA binding motifs using MEME-suite.

However, recently I re-downloaded non-redundant CORE PFMs in MEME format from JASPAR website and found the error occured during analysis:

Errors from MEME text parser:

The PSPM of motif MA0847.2 has probabilities which don't sum to 1 on row 2.

FATAL: An error occurred reading the motif file.

The meme file of MA0847.2.FOXD2 looks this:

MOTIF MA0847.2 FOXD2

letter-probability matrix: alength= 4 w= 13 nsites= 5994 E= 0

 0.234234  0.305973  0.386053  0.073740

 0.144311  0.401401  0.051552  0.598765

 0.364198  0.006006  0.000501  0.635969

 1.000167  0.015349  0.215883  0.003837

 0.414081  0.201535  0.125459  0.259092

 0.435102  0.016850  1.000167  0.001001

 0.024691  0.167668  0.000000  1.000167

 1.000167  0.109776  0.001835  0.001835

 1.000167  0.034868  0.002836  0.001835

 1.000167  0.000000  0.004171  0.005506

 0.000000  1.000167  0.000000  0.383217

 1.000167  0.008842  0.020854  0.034868

 0.400400  0.167334  0.085919  0.346346

URL http://jaspar.genereg.net/matrix/MA0847.2

Obviously, there are rows whose sum > 1 and even one of the frequency is larger than 1.

MEME-suite clearly documented that 'the frequencies should sum to 1'.

What happend to this PFMs?

Best,

Byunghee Kang

[Beisi Xu]

I encounter the same problem. For https://jaspar.genereg.net/matrix/MA0841.1/ 

MEME version 4

ALPHABET= ACGT

strands: + -

Background letter frequencies

A 0.25 C 0.25 G 0.25 T 0.25

MOTIF MA0841.1 NFE2

letter-probability matrix: alength= 4 w= 11 nsites= 18831 E= 0

 0.296798  0.382773  0.281610  0.038819

 1.000000  0.000478  0.069354  0.000266

 0.000106  0.000425  0.000000  1.000000

 0.000000  0.000000  1.000000  0.000106

 1.000000  0.000584  0.000000  0.000212

 0.001965  0.762572  0.237481  0.000903

 0.000000  0.001062  0.000159  1.000000

 0.000053  1.000000  0.000000  0.000159

 1.000000  0.000690  0.000425  0.000266

 0.000000  0.142796  0.000531  1.000000

 0.025862  0.523605  0.269078  0.181456

URL http://jaspar.genereg.net/matrix/MA0841.1

[Oriol Fornés]

Sorry for the late reply.

There seems to be a problem with how the probabilities are being computed.

Apparently, it applies the sum of the first column to every column.

I have developed the following script that translates a JASPAR profile into several formats such as MEME.

I hope it helps.