How To Run The Meme Motif Discovery Software On A Large Dataset?

Question

[k.nirmalraman]

I am currently using MEME for Motif Discovery and I would like to check about 50 to 100 bases upstream for binding factors (say they are represented around -35 and -10 usually). I have a local installation of MEME.

I have about 30K upstream sequences and I am not able to run the algorithm even with -maxsize set to any high values and I get

Error: Dataset too large (-1) Rerun with larger -maxsize

How can I address this problem?

Also, as an extension to this question:

I am expecting to find more than one motif conserved (say at both -35 & -10) in different subsets of the 30K Sequence. How can I make such specifications (of location range of motif) while running MEME? or is there a variant of MEME that does this particularly?

As much as I understood PSP file, I am not able to understand what exactly does bgfile do in MEME motif discovery?

Thanks!

[Niallhaslam]

Can you post the options you are using to run MEME?

Answer 1

[Mikael Huss]

The simplest solution may be to run DREME instead. It's available from the same website (because it was developed by the same group) and made with larger (ChIP-seq scale) data sets in mind. There are also many other tools that you can try for your data size. MEME simply doesn't scale that well.

[k.nirmalraman]

Thank you for the suggestion. I shall try DREME!

[k.nirmalraman]

sure!

meme ~f1.fasta -dna -mod oops -w 8 -minw 6 -maxw 8 -nmotifs 5 -psp dna4_8.psp -revcomp -maxsize 1000000000000 -o ~/MEME/

[a1ultima]

Hi @Michael-Huss (@Michael Huss) cheers +1, do you know if DREME allows parallel processing?