Give different PCR cycles of samples in one CUT & TAG library preparation

Question

[sir.outman]

Dear fellows:

Here is the background of my questions: During the CUT & TAG library preparation, we found the library concentration of some samples is low, and can not match the required concentration of the sequence factory. so we give some extra PCR cycles to low concentration(even different cycles between the replicates of the sample), the PCR cycles are ranged from 15 to 18 depending on their DNA concentration. After this, the sequence factory performed Cut & TAG seq. The duplicate rates got from MultiQC report range from 47% to 76%.

Here are the questions:

1. Give different PCR cycles(extra PCR cycles for low concentration samples) instead of giving the same cycle for all samples to match the sequence factory requires, does this make sense?
2. Will those extra PCR circles cover the biological difference when performing Differential Binding Analysis between samples?
3. Will different cycles between replicates pull down common peak numbers that appear in replicates?
4. According to the protocol from Steven Henikoff's lab, they said "extra PCR cycles reduce the complexity of the library and may result in an unacceptable level of PCR duplicates", are our data's duplicate rates(47% to 76%) acceptable?

Thanks for any advice!

[ATpoint]

1) If you need more cycles for some samples then go ahead, I mean, what is the alternative? 2) In my experience, no. A few PCR cycles should not be a major confounder. 3) Not sure what you mean, if you want reproducible peaks then you can intersect peak sets between replicates or use frameworks such as IDR to automate that. 4) Again, if you need extra cycles to get enough material to sequence the sample then go ahead. 76% is a lot, if you remove the duplicates you have to see whether the remaining signal is still usable, do PCA and other common diagnostics to see whether the particular samples still cluster with the samples of that group and somewhat away from other groups. Check if number of peaks is somewhat near to the number of the replicates. There are no fixed rules for all this, it really depends on the dataset.

[sir.outman]

Hi @ATpoint ,

Thank you so much for the reply, may I continue to ask more on this question's thread?

I'm worried about if we give different PCR cycle numbers between conditions but instead of giving them the same number of cycles, like give 15 cycles to condition_A and give 18 cycles to condition_B, when performing Different binding Analyses of CUT&TAG data, will those unequal PCR cycles(3 extra cycles on condition_B) bias the results? Did the step of removing duplicate reads can some kind reduce this bias?

[ATpoint]

I cannot say that I ever observed confounding by PCR cycles, at least not in a PCA which is usually what I do to check for any signs of confounding. I mean, you have to do as many cycles as necessary to have enough material, so there is not really a way around that.

[sir.outman]

Got that. Thank you so much!