Entering edit mode
8.1 years ago
Sudhir Jadhao
▴
70
Hello Everyone,
I have Macs 1.4 result. I want to filtered out MACS result on the basis of NSC and RSC score.
Can anyone please tell me , how I should proceed for the same ?
Than you Inadvance.
NSC and RSC are sample-level metrics, so the filtering would simply be ignoring entire files.
Yes these metrics are obtained on your samples with SPP tool as well to test the quality of your ChIP-Seq data. In that case discard sample having the flag of -1 or -2 that are coming from the output of SPP. So filtering the entire sample. To strengthen your selection you can always use CHANCE to couple with SPP output and see which samples have low enrichment against background and discard them if the same samples are also listed to be of poor quality with SPP. These are QC tools for the aligned samples. For filtering on the peak files it is advisable to do it with FDR or value and fold enrichment scores.
P.S. : Why MACS1.4 still now? MACS2.1 is up and running and then there are more peak callers out in the field why stick to the old tool. Just a piece of advice. Upgrade your tool.
@ vchris_ngs Thank you for your replay.
So you want to say NSC and RSC are the score , which we use to see how our alignment is? Is there any corelation between NSC and RSC score with FDR ?
If NSC and RSC score are less than 1.1 and 0.8, we discard whole sample?
There is no relation to peak calling. These scores are trying to give information about the ghost peak or rather say that the quality of your chip is not that great if your fragment length is dominated by the read length. However when you use the SPP scores you have also a score ranging from -2 to 2. Taking a look at those scores one can discard samples with -1 and -2 scores since they are poor quality samples meaning the quality of the chip is not that great. These are all performed before peak calling. Nothing to do with any FDR metrics of peaks.
Take a look at this link for more information.
@Devon Ryan
Thank you for replay..
so NSC and RSC score are calculated before peak calling ?
Yes, the NSC and RSC scores have nothing to do with peak calling, in fact peakier parts of the signal tend to get excluded (calling the algorithm that's used to calculate these "weird" is an understatement). As an aside, I find these scores largely meaningless.