MEME results are recorded in three file formats:
The MEME HTML format is meant to be easy to read and provide useful visualizations and tools for continuing to work with the discovered motifs. The HTML output is generated from the XML output by the use of a XML stylesheet transformation.
The MEME XML format is completely specified by the Document Type Definition (DTD) found at the start of the MEME XML output. The XML format is intended for machine reading as it is just the data without any explanatory text or visualizations. The XML format was added for MEME 4.0.
The MEME plain text format was the original output from MEME and like the HTML contains explanatory text making it self-documenting. The minimal MEME format is essentially the MEME text format without the documentation.
GLAM2 results are recorded in two file formats:
The GLAM2 format is described in the Output format section of the GLAM2 Tutorial. GLAM2 also provides a MEME minimal motif format approximation of the motif.
Most MEME Suite programs, which accept motifs, can use the outputs of MEME and DREME (plain text, HTML and XML) as well as the MEME minimal motif format.
The major programs which accept MEME and DREME output are:
GLAM2SCAN will only accept the plain text and HTML forms of GLAM2 output.
Users may create motif files in a simplified format for use by the MEME Suite programs (excluding GLAM2SCAN).
Prior to release 4.7.0 of the MEME Suite MAST required a log-odds matrix section to be specified, however the current version of the MEME Suite is capable of translating the letter-probability matrix into the log-odds matrix and vice-versa.
The Minimal MEME format contains following sections:
For each motif in the motifs section there are the sub-sections:
*Note that at least one of the two starred sections is required.
MEME requires this line to be certain that it really is reading a MEME motif file and not just something that looks slightly like it. This line must appear before any other sections in the file.
The version number should be the MEME Suite version you are targeting.
For example to target MEME Suite version 4 and above:
The alphabet line tells the MEME Suite what alphabet to expect the motifs to be in. If this line is not present then the MEME Suite can attempt to detect this from the background or the motifs themselves.
The alphabet can be ACGT for DNA or ACDEFGHIKLMNPQRSTVWY for protein.
For example using a DNA alphabet:
The strands line only has meaning for DNA motifs and indicates if motifs were created from sites on both the given and the reverse complement strands of the DNA sequences. If this line is not supplied then the MEME Suite will assume that DNA motifs were created from both strands.
The which strands can be replaced with + to indicate only the given strand and + – to indicate both strands.
For example to indicate only the given strand was used:
The background frequencies tell the MEME Suite how prevalent each letter of the motif alphabet was in the source sequences which were used to create the motifs. If the background frequencies are not supplied then the MEME Suite will assume uniform background frequencies. The MEME Suite uses this background to convert between motif letter-probability matrices and log-odds matrices.
The source is not required and if you wish you can leave off then end of the first line after “Background letter frequencies”. On the next line is listed each letter in the alphabet followed by its frequency. The letters must be listed in the same order as in the alphabet line and the frequencies should sum to 1.
An example of uniform DNA frequencies:
An example of protein frequencies with a source listed:
The motif name line indicates the start of a new motif and designates an identifier for it which much be unique to the file. It also allows for an alternate name which does not have to be unique.
For example:
The letter probability matrix is a table of probabilities where the rows are positions in the motif and the columns are letters in the alphabet. The columns are ordered alphabetically so for DNA the first column is A, the second is C, the third is G and the last is T. For protein motifs the columns come in the order A, C, D, E, F, G, H, I, K, L, M, N, P, Q, R, S, T, V, W and Y. As each row contains the probability of each letter in the alphabet the probabilities in the row must sum to 1. If this section is not specified then the log-odds matrix must be specified.
All the “key= value” pairs after the “letter-probability matrix:” text are optional. The “alength= alphabet length” and “w= motif length” can be derived from the matrix if they are not specified, provided there is an empty line following the letter probability matrix. The “nsites= source sites” will default to 20 if it is not provided and the “E= source E-value” will default to zero. The source sites is used to apply pseudocounts to the motif and the source E-value is used for filtering the motifs input to some MEME Suite programs (see MAST’s -mev option).
An example of a DNA motif’s letter-probability matrix:
If you’ve included the letter-probability matrix then unless you want MAST to use a specific log-odds matrix then there is no reason to include this section. The original version of MEME only output motifs with a log-odds matrix. Later on the letter-probability matrix was added and MEME and MAST were merged with Meta-MEME to become the MEME Suite. Since then all new programs have been written to use the letter-probability matrix leaving MAST as the only program which uses it. In the output from MEME the log-odds matrix has some additional tweaks (especially in protein motifs) which are impossible to perform with just the letter-probability matrix and the 0-order background so it has been kept around.
The log-odds matrix is a table of scores where the rows are positions in the motif and the columns are letters in the alphabet.
The scores are calculated as follows:
let bL be the background probability for letter LAs the log of 0 is negative infinity then pseudocounts should be added to the probabilities first.
let pLi be the probability for the letter L at position i in the motif
let sLi be the score for the letter L at position i in the motif
sLi = round((log(pLi / bL) / log(2)) * 100)
As in the letter-probability matrix the “key= value” sections are optional and have the same defaults
An example of a DNA motif’s log-odds matrix:
The URL line specifies a web-page to link to when mentioning the motif in results.
For example: