Merging MACS Peak analysis +/- duplicates?

[Combo]

My ChIP-Seq data has a relatively high (40-85%) number of duplicate reads. This presents problems for Peak calling.

Approach 1. Using MACS, I can analyze the data and choose to count duplicate reads as a single read. This produces approx. 5000 peaks and the data looks reasonable, however some important/known occupancy regions are absent.

Approach 2. Alternatively I can use MACS and specify 'KEEP ALL' for the duplicates. Naturally this means the peaks in the first approach become dwarfed by the large duplicate peaks, so a completely different set of peaks are identified.

This approach produces approx. 100,000 peaks (depending on other settings) including the important occupancy region we are expecting which also has a very high MACS score (approx 75 percentile within this dataset).


What would be the best way to proceed?

Would it be reasonable to filter the 100,000+ peak data with a MACS score threshold? Would there be a statistically valid (i.e. non-arbitrary) way to do this? e.g. converting the MACS scores to p-values and using an FDR cut-off?

Would it be reasonable to merge the two datasets? i.e. the peaks identified with the duplicate reads and those identified without.

[ETHANol]

Ultimately qPCR will tell you what method is working and what is not. That being said, real peaks look different to the human eye then peaks that arise from PCR amplification.

You also need to repeat the experiment. There are a lot of papers published with just one experiment/one antibody, but it is sloppy science. Hopefully, sooner rather then later the bar will be raised.

[Combo]

You're right I would much prefer to have replicates. Unfortunately this is all I have to work with for the time-being.

Any thoughts on merging the dup + dedup peak analyses?

Also is there a statistically valid way to apply a MACS score threshold cutoff?

PS - So far I've been running MACS online via Galaxy. Any ideas on how fast it would run locally on a standard desktop (win 7 / 2gb ram / pentium core 2 duo)? I am curious about using the advanced command line options to play with a manual duplicate limit.

[ETHANol]

You can merge if you like or not merge. The only way to tell if your peaks represent in vivo binding sites is to validate by qPCR. Design some oligos to some peaks MACS has identified (and some places where it looks like it isn't biding) and see if your peaks are actually biding sites by ChIP-qPCR. You have to do this before proceeding regardless. Don't waste your time on further analysis or experiments until you have verified your peaks by ChIP-qPCR.

Another thing to think about is garbage in equals garbage out. Yes, I know you want to get something out of your data, but your better off repeating your experiment.

As far as running MACS locally on your computer, you can probably do it with 2GB of RAM but that is going to be borderline. Your processor speed doesn't really matter (just slows things down). Not enough RAM and it crashes. Upgrade your RAM if you can.

[Joe Petrosino]

Hi all,

I started to work MACs by Galaxy. I read on "Readme for MACS" that: "For the experiment with several replicates, it is recommended to concatenate several ChIP-seq treatment files into a single file".
Now, I have illumina ChipSeq data: two files for IP samples and two files for Control samples. Is It right to use Concatenate datasets (text manipulation) and then use MACS for the peaks calling?
Furthermore, performing MACs peak calling, even after lots of parameter optimization, the results shows an FDR > 14%. Isn't it very strange?
Any suggestions are very welcome.
thanks!

[mudshark]

@Combo:

do you have a control sample (input or IgG control)?
how many reads in total do you have?
which is your model organism?
what is your target (i.e. txn factor, histone modification..)?

[Joe Petrosino]

Hi Mudshark,

1)I have illumina ChipSeq data: two files for IP samples and two files for Control samples.
2)I have about 110.000.000 lines after Map with Bowtie for each files
3)Mouse
4)Transcription factor

Is It right to use Concatenate datasets (text manipulation) and then use MACS for the peaks calling?

Thanks

[mudshark]

Hi Mudshark,

1)I have illumina ChipSeq data: two files for IP samples and two files for Control samples.

are these biological replicates? do you have the same duplication levels in the input sample?

2)I have about 110.000.000 lines after Map with Bowtie for each files
3)Mouse
4)Transcription factor

if you have a lot of mapped reads (as you have) you might actually run into a problem when you de-duplicate as the probability of getting 'true' duplicates increases with sequencing depth. The 'true' duplicates will most likely map to your true binding regions.

Is It right to use Concatenate datasets (text manipulation) and then use MACS for the peaks calling?

Thanks

if your files reflect technical replicates you can concatenate them. if not (i.e. they are biological replicates) I would strongly suggest individual processing.

You might want to try the new version of CisGenome which is the only software (afaik) that can deal with replicates and does so quite convincingly.

[Joe Petrosino]

Hi murdshark,

My files are technical replicates...I'll also try CisGenome.
Thanks for the quick answer!