Large Number Of Macs Negative Peaks

Question

[Nick]

I have a TF chip-seq study with two types of replicates:

• 2 Flag+TF samples, each with a corresponding IP control (Type A)

• 2 Flag- samples, each with a corresponding IP control (Type B)

I am attaching stats (% uniquely aligned, non-aligned, duplicates, initial number of paired peaks, estimated fragment size, number of positive and negative peaks) about each treatment-control pair:

As you can see, the type B treatment-control pairs have a large number of negative peaks (9511 for the treatment-control pair in the first table row and 2009 for the treatment-control pair in the third row). I would not have expected any peaks - neither positive nor negative as no TF factor was added to any of these samples. There are some other anomalies (e.g. high percentage of non-aligned reads in the IP controls) which may or may not be related to this phenomenon. I checked some of the non-aligned reads but they don't seem to produce any BLAST hits so they may be some sort of technical artifacts. I don't know how to proceed with these samples - shall I throw them out? Any advice is most welcome.

[Ulduz]

I am having similar problem with my some of my samples (more negative peaks than positive peaks), I am curious to know how did you proceed?

Answer 1

[Lorena Pantano]

I would remove duplicated reads, you seems to have a lot in the negative control (17%), then run again and see what happens.

Also you can check the region of some of the duplicated reads are, to see if they are exactly the same or what, normally theyr are artifact of PCR.

Any difference of the total number of reads, not %, total number of reads mapped in the control and the TF experiment?