How to run Meme on cluster

The latest version of Meme, v4.6.1, was built with mpich, while the earlier versions were built with openmpi. You need to set the path correctly as in the examples below.

Your input database should consist of a file containing sequences in fasta format. In the example below, the file is 'mini-drosoph.s'.
Maxsize parameter: The maximum dataset size in characters. Determine the number of characters in your dataset by typing 'wc -c filename'. e.g.

[user@biowulf mydir]$ wc -c mini-drosoph.s 
506016 mini-drosoph.s

For this dataset, the maxsize parameter has to be set to greater than 506,016, so we will use 600000.Set up a batch script along the lines of the ones below:

Batch script for Meme 4.6.1

----  this file is called meme.batch ---------
#!/bin/bash
#This batch script can be used with Meme 4.6.1
#PBS -N Meme
#PBS -m be
#PBS -j oe

export PATH=/usr/local/mpich-1.2.7p1-gcc4_64/bin:$PATH
cd /data/user/meme/
time mpirun -machinefile $PBS_NODEFILE -np $np /usr/local/meme_4.6.1/bin/meme_p \
     /data/user/meme/test.fa -oc /data/user/meme/meme_out \
     -maxsize 10000000 -p $np

// The above tips are from http://biowulf.nih.gov/apps/meme.html

btw, the above script use PBS (Portable Batch System) submission system. Alternatively, there are other options, like LoadLeveler (Champion), LSF (Lonestar). There are syntax comparison here. Also, more detail for SGE and PBS here. 

Difference between FIMO and MAST

FIMO and MAST are two applications in the MEME suites.

FIMO is looking for the best individual matches to motifs. MAST is looking for sequences the have the best overall match to a collection of motifs.

FIMO's task is simple: given a set of motifs and a database of sequences, compute the match score to each motif at each position in each sequence, and report all the motif matches that pass the p-value/q-value threshold.

MAST's algorithm is more complex. For each sequence it carries out an initial scoring that is quite similar to FIMO's. MAST then picks the best match for each motif in the sequence. The p-values for these top matches are multiplied together to create an overall score for the full sequence. MAST reports the sequences that have the most significant overall scores. Typically MAST would be used to look for regulatory regions in DNA, or structures in proteins, where several motifs might occur near each other.

-- Thanks to Charles for the explanation