Correlation of TF Binding with expression data

Question

[Prakash]

Hi,

I would like to perform correlation analysis between gene expression and TF binding in control and KD condition. So what I have till now is gene expression in control and KD condition and normalized tag count of different transcription factor obtained from Homer annotatePeaks.pl for merged peaks of all the TF. One approach I am thinking to take average tag count for all the genes and correlated with expression or another way correlate fold change of TF binding with fold change in expression. Do you think its a right way to do this analysis ? If not, please suggest some approach or any tools that perform this type of analysis. Any comments or suggestions will be highly appreciated.

Answer 1

[venu]

Do you have TF ChIP-seq data or Histone ChIP-seq data and potential TF binding sites?

Since you did not mention about that, I would assume you have TF ChIP-seq data, one direct approach would be,

1. Assign each peak/TFBS to nearest 1 or 2 genes (1 upstream and 1 downstream of peak)
2. Calculate normalized tagcount/enrichment for each of the binding site from step. 1
3. Perform correlation between each TFBS tagcount with the expression across all samples

This will give an idea about which TFBS are highly correlated with genes. This gets very complicated for Histone ChIP-seq data.

[Prakash]

Thanks Venu !

I have ChIPseq data for TF in control and KD condition. I have annotated the TFBS/peak to nearest genes using homer (default it annotates to only one nearest gene) and have normalized tag count as well. Each binding site can be correlated with expression, but what you would do if you have multiple TFBS to one gene and how would you correlate each binding site with expression across multiple condition.

[venu]

if you have multiple TFBS to one gene and how would you correlate each binding site with expression across multiple condition

It's not a problem unless TFBS are overlapping themselves. I don't know what is your actual goal. But I would simply do each TFBS to GENE correlation and take the highest/lowest correlated one depending on the goal.