Issue with replicating from a research paper-motifs

Question

[joshua.steier]

Greetings, I've been trying to replicate results from the following research article: https://bmcbioinformatics.biomedcentral.com/articles/10.1186/s12859-018-2378-9 In the paper, there is a figure that describes motif #57 with a sequence logo and its position weight matrix.

I took the data from the given datasets in additional material section and gathered the FASTA files needed to process through MEME.

The combined dataset(denoted CM in the paper) was processed through MEME but I obtained different motifs than those obtained by the authors. I checked the datasets and my FASTA files and they are correct, I'm unsure of how this discrepancy between my results and theirs exists, as I followed the paper.

I suspect that maybe the FASTA files changed over a year in the Uniprot database since the paper is from 2018 and we are in 2019. I searched in MEME for motifs with length 3 to 10, as this describes the motifs discussed in the paper.

Any thoughts or possible directions?

Here is the Motif they are showing: (57)

Here is the motif I got:

[Philipp Bayer]

Have you checked whether you're using the same MAST/MEME versions? I've seen quite a bit of discrepancy there

[joshua.steier]

You're correct to address that, I believe our MEME versions are different. The one I'm using is 5.0.5 and the one they used is most likely 5.0.0. Would the difference in version cause this much of a difference?

[Philipp Bayer]

I've had some large differences in the past, but with earlier versions than that - have you tried rerunning with MAST/MEME 5.0.0 from http://meme-suite.org/meme-software/5.0.0/meme_5.0.0.tar.gz ?

[joshua.steier]

Hi, I just ran MEME under that version for the motifs and got the same results as I got before.

[Philipp Bayer]

Thanks for checking :)

[joshua.steier]

Thank you for your help :)

[joshua.steier]

Update: I found out the authors are using MEME version 4.10.1. Do you think this likely caused the discrepancy? I'd like to test it out but I'm having some difficulty getting 4.10.1 to work on linux server.

[Philipp Bayer]

Interesting! Conda has 4.11 and 4.8, maybe try those?

https://bioconda.github.io/recipes/meme/README.html

You can either install the version as a Docker image:

docker pull quay.io/biocontainers/meme:4.11.1-0

or via conda:

conda install meme=4.11.1-0

replace 4.11.1-0 by 4.8.1-1 to see the really old meme - if you see the same results as yours' in 4.11 and 4.8 I'd doubt they appear in 4.10

[joshua.steier]

Thanks! Using Conda was much better than the standard methods. I ran it on 4.11 and 4.8, the authors' results showed on both of those. So the version must be causing the discrepancy as you said. I'm going to try to replicate given the now correct motifs from the old version. Do you suppose it is also worthwhile to replicate their methods using the new motifs from the newest MEME version?

[Philipp Bayer]

Interesting!! Good to know that you got the old result :) If the motifs are so different, which one is the 'correct' version then? I presume their results will not hold up using the newest MEME version, or you'll get completely different results..

[GouthamAtla]

Can you elaborate/show which motif they are showing and which motif you got ?

[joshua.steier]

Here is the Motif they are showing: (57)

Here is the motif I got:

[GouthamAtla]

Thanks. I added it to the question. Looks like a huge difference. Did you check all our motifs ? May be not the top motif but the motif authors showed might be somewhere on the list. If not, I would directly contact authors.

[joshua.steier]

Thanks for adding it to the question, your help is greatly appreciated. I looked at my results again and couldn't find that motif. I'll try contacting the authors, I appreciate your suggestions.