I'm comparing with SICER some chip seq experiments of mouse muscles subjected to a specific treatment vs non treated samples.
I'm monitoring ACH3 and H3K4ME3 histone modifications(so they are both activator at promoter level).
The strange thing is that most of the genes that are modulated according to their gene expression have at their promoter increased ACH3 islands and decreased H3K4ME3.
Do someone of you experienced a similar situation?
This could be a signal of something wrong in experiment setup?
I find that these kinds of issues are always the hardest to answer! I have never used SICER but here are some very general strategies that I have found to be helpful when attempting to determine whether a strange result is real:
Manually check a few genomic locations that you know the answer too (from literature or previous work). I.E. is there a location that you know should have increased ACH3/decreased H3K4ME3, or you know should exhibit the opposite. Try visualizing the called peaks and even the raw reads as wig tracks and see if you can find validate some small sites
Do some QC analysis of your data. Run your raw FASTQ reads through FASTQC and see how they score. Things like PCA analyses can be useful in spotting outliers in your data at the aligned read stage. I have personally found this to be very important in differential/comparison experiments with replicates. Sometimes one replicate of a condition will be way off the mark and will do strange things to the overall comparison between the groups
Re-do the analysis with a different tool/software. I have used MOSAiCS for calling doing CHiP seq work. Give that a try and see if you get the same answer
Lastly, see if you can validate this phenomena using wet lab techniques. I can't offer advice on that, but it has proven to be a useful tool for me when collaborating with biologists!
