Difference in signal-to-noise in MNase-Seq at 100 vs 200 vs 400 million reads

Question

[cpak1981]

Hi,

Has anyone determined how signal-to-noise changes in MNase-Seq profiles with 100 vs 200 vs 400 million reads? Does it improve? Or does it stay the same regardless?

Thanks!

[Carlo Yague]

How do you define noise in this context ?

MNase-seq is about levels of MNase protection and it depends on many experimental things but not read depth IMO. Increasing depth will only gives you a more precise measure of the MNase resistance.

[cpak1981]

Thanks for the follow up question and answer.

In terms of signal-to-noise, I guess I am actually asking about base-pair resolution of nucleosome positioning vs sequencing depth. I framed my question as signal-to-noise because assuming Gaussian-like distributions the "resolution" of Gaussian-like profiles can be expressed as the full-width half-maximum (FWHM), which depends on the peak height (aka signal) and variance (aka noise).

I'm looking into this issue now but don't really have a full answer at the moment.

[GenoMax]

Please use ADD REPLY/ADD COMMENT when responding to existing questions to keep threads logically organized

[Carlo Yague]

The reads that fall between the peaks are not "noise" in my opinion. These are reads that reflects partial protection to the MNase. This can be due to incomplete digestion, weak nucleosome positioning within the popultion, TF binding sites,... That is why I asked you what you mean by "noise".

So depth will not change the ratio between the signal in the peaks and the signal between the peaks. Depth is important in order to see those signal of course, but the ratio remains the same.

To optimize the ratio (since this seems to be your goal), you need to optimize the experimental conditions (digestion time, ...). Also paired-end sequencing (vs single) helps to get precise peak location.

Answer 1

[Carlo Yague]

This method paper for human MNase-seq recommends >10x coverage. In this other paper on yeast, they reach between 13x and 177x fold coverage and consider it sufficient.

So I would say that at least 10x coverage is enough for proper peak definition.

Answer 2

[cpak1981]

I agree that reads that fall between the peaks are not "noise" per se, and that these reads reflect partial protection due to various factors. But sequencing depth certainly has some influence on peak height (signal at peaks) and peak width (signal between the peaks). I don't think there's a debate that mapping 100 individual reads (for an individual nucleosome) will give a "more Gaussian-looking" profile than 10 individual reads; in other words, peak height will improve with read depth (up to a point of course). And that 100 individual reads will give a better estimate of the "average" nucleosome position than 10 individual reads. I agree with your advice that more experiments under various conditions would help inform on nucleosome positioning, but even under these different questions, I'd still like to know: For a given library, at what read depth does the ratio of peak height and peak width start to be the same?