Tweet
I have 2 batches of chip-seq samples:
(A) One biological SE replicate
This batch, actually, consists of 4 SE Chip-seq samples - one treatment, one control, both have IP controls. All are SE sequenced at a lab A
(B) Two biological PE replicates
This batch consists of 8 PE Chip-Seq samples - 2 treatment, 2 controls, all have IP controls. All are PE sequenced at a lab B.
The idea is to establish differential binding between the treatment and the controls. I plan to use macs and diffbind but I am open to suggestions.
My first specific question is how to treat the PE samples. I see 3 options:
(1) Treat all forward and reverse reads as SE reads
(2) Take just the forward or the reverse reads
(3) Treat each PE sample as, essentially, 2 technical replicates (one consisting of forward reads, one of reverse)
I find (3) intuitively attractive but, again, I am open to suggestions. If I go for it than the model would need to introduce 2 additional factors - one accounting for the technical replicates, another for the batch effect. So far I have dealt with batch effects only - I presume that adding additional factor shall be straight-forward but, please, let me know if there are things that I need to pay attention to.
(A) One biological SE replicate
This batch, actually, consists of 4 SE Chip-seq samples - one treatment, one control, both have IP controls. All are SE sequenced at a lab A
(B) Two biological PE replicates
This batch consists of 8 PE Chip-Seq samples - 2 treatment, 2 controls, all have IP controls. All are PE sequenced at a lab B.
The idea is to establish differential binding between the treatment and the controls. I plan to use macs and diffbind but I am open to suggestions.
My first specific question is how to treat the PE samples. I see 3 options:
(1) Treat all forward and reverse reads as SE reads
(2) Take just the forward or the reverse reads
(3) Treat each PE sample as, essentially, 2 technical replicates (one consisting of forward reads, one of reverse)
I find (3) intuitively attractive but, again, I am open to suggestions. If I go for it than the model would need to introduce 2 additional factors - one accounting for the technical replicates, another for the batch effect. So far I have dealt with batch effects only - I presume that adding additional factor shall be straight-forward but, please, let me know if there are things that I need to pay attention to.