enrichment\depletion

[mariama...@gmail.com]

Hello,

I have a question regarding motif search.

I compared 10 chip-seq data set and I clustered them and I got 15 different clusters. I want to do a motif enrichment for known motif on each cluster and then compare the enrchemt of each motif across all samples. I thought I can use either Centrimo or ame to see enriched/depleted motifs.

But If I run 15 different clusters with ame/Centrimo as 15 different jobs and I get some motif , how should I compare the enrichment values for individual motifs among other clusters (e.g . I know that e.g. in cluster1 motif X has a certain p-value but in cluster 2 this motif get another score)? results from different jobs are directly comparable or I should do something else? should I use the same background or control e.g pool all 15 clusters and upload it as control seq?

thank you so much in advance

Best

M

[CharlesEGrant]

p-values should be directly comparable across runs. Be sure to use the adjusted p-value to reduce multiple-testing issues.

I think using the same background and control sequences for each run is probably a good idea, but I wouldn’t just pool all 15 clusters to create the control sequences. The ideal control sequences are ones that you know are not enriched for the motifs. You could pool all 15 clusters, but then you should shuffle them, so that any matches to your motif are purely by chance. AME generates the default control sequences by shuffling the target sequences, but if you provide your own control sequences AME doesn’t shuffle them.