I would like to plot a heatmap of ChIP-seq binding sites among 3 transcription factors. The only normalization I did now is log2 ratio to the inputs with depth normalization (SES in Deeptools). Is there any other normalization should be done?
Here is the experimental design:
ChIP-seqs on: TF1, TF2, TF3, Input
In each genotype: WT, TF1-knock down (KD), TF2-KD, TF3-KD
Each has 3 replicates, so total bam files: 4x4x3=48.
The purpose is to identify binding sites that co-regulated or independently regulated by TF1, TF2 or TF3.
I saw some post is suggesting IDR: How different the different ChIP-seq samples are .... and scale all datasets library size: How To Do Normalization Of Two Chip-Seq Data and Comparing numbers of ChIP-seq peaks between different sample types.
But these are mainly compare among samples. If I want to compare among TFs, any other normalization I should do? (antibody efficiency or other biases?)
For the heatmap plot, should I get overlapped peaks first and then plot heatmap or plot over all union of peaks and using some clustering method to identify overlapped peaks?
Now I plotted read counts for all union of peaks among 3 TFs as heatmap with Hierarchical clustering, but I'm concerning the read counts enrichment is not being normalized.
Any suggestions on other post links, publications, and softwares would be appreciated! Thank you all in advance for your advises and suggestions!!