Homer motif analysis: "Very close % of target vs % of Background sequences with Motif" & "SeqBias at 1st rank"

Question

[chiefcat]

Hi everyone. I'm trying to find the motifs bound by my ChIP-ed protein using Homer. I don't have knowledge regarding NGS analysis. The analysis was done by my labmate who understand programming languages but lack of experience in analysing ChIP-seq.

When going through the result list, I found that there are 2 obvious issues happened.

1) The % of target vs % of Background sequences with Motif are very close throughout the list. When this happened, does it mean that the motifs found are actually not/ very lowly enriched in my ChIP-ed sample?

2) The "SeqBias: CG bias" appears as the 1st ranked motif in my list. I know this indicates that the analysis parameter is not appropriate as stated in http://homer.ucsd.edu/homer/motif/practicalTips.html. This is probably related to the normalization to either by total GC-content or CpG-content if I understand the tutorial correctly. Should I just adding "-gc" to the motif finding command to solve this problem? If having this problem on my list, do other results on the list still valid?

Thanks very much!

[os306]

Almost six years later, I am facing a similar situation when interpreting my ChIP-seq data. I was wondering if you were able to figure out/troubleshoot what the issue was?

Thank you!

Answer 1

[ildem]

Hi, I usually get a decent enrichment over background, sometimes upto 3 fold, but at least a 1.5 fold would be expected. Can you give us your command line? What does your bedfile (input) look line? Could it be that you are searching for a motif in too large of an area? Usually you search for the motif within +/-100 bp of the peak summit, and +/-250bp for the "co-binding" Tfs. If you search for a motif in large areas (i.e. entire peak) you might end up with high background.

Also, Have you removed the "blacklisted" regions from your Chip-seq peaks to get rid of the background regions that are called peaks by every single ChIP experiment (type blacklisted regions and your genome in google).

As Homer says, I would try to match the background to your Chipseq peaks. Are they mostly promoters? then generate a background file for promoters etc.. by default homer is using randomly picked regions as background.