Treatment of ChIP-Seq Input
Hello all! I have placed this in the Bioinformatics forum because I believe this might be a more statistical problem, but if I'm incorrect, please don't hesitate to let me know.
Briefly, I am doing a large (for me) ChIP-seq project. I have two species across three developmental stages. Each stage has condition 1 and condition 2, and there are replicates of each. Every sample had their IP and Input sequences to a minimum depth of 20M. (48 libraries: 24 Input, 24 IP).
My question is about how to treat my inputs. Obviously, it would be ideal (I think) to normalize each IP against it's specific Input. That would be the most specific normalization.
However, I can also imagine that by pooling the Inputs within a treatment (but not across replicates or species), I could compare each developmental stage within a treatment to a common Input with 3x greater depth.
A single replicate example is below (there would be two replicated datasets like this)
Any thoughts as two which would give the most robust results? I'm planning on running it both ways, but was hoping for some insight from someone with more experience in analyzing this type of data.
Best and many thanks,
Bob
Briefly, I am doing a large (for me) ChIP-seq project. I have two species across three developmental stages. Each stage has condition 1 and condition 2, and there are replicates of each. Every sample had their IP and Input sequences to a minimum depth of 20M. (48 libraries: 24 Input, 24 IP).
My question is about how to treat my inputs. Obviously, it would be ideal (I think) to normalize each IP against it's specific Input. That would be the most specific normalization.
However, I can also imagine that by pooling the Inputs within a treatment (but not across replicates or species), I could compare each developmental stage within a treatment to a common Input with 3x greater depth.
A single replicate example is below (there would be two replicated datasets like this)
Any thoughts as two which would give the most robust results? I'm planning on running it both ways, but was hoping for some insight from someone with more experience in analyzing this type of data.
Best and many thanks,
Bob