Help with picking an even sampling depth

Hi dear all,
I am currently practicing with QIIME2 tutorials. I was able to proceed with the Moving-pictures-tutorial. However, when I try with the fecal microbiome transplant study tutorial (1% subsample data), I had the problem with sampling depth. Initially, I entered 1109, but there was an error. Thereafter, I reviewed the information presented in “table.qzv” and select the value as 500 although the maximum value is 860. Based on the moving-pictures-tutorial, it indicated that choose a value that is as high as possible to retain more sequences per sample. I am wondering whether I have selected it correctly.
And also, I had confusion that when I review the sampling depth in “table.qzv”, I increased the sampling depth while checking the retained sequences number (%). It showed me an inverse relationship between sampling depth and retained sequence number (%). In this case, is it possible to select a sampling depth value as high as possible?
Could you please be kind enough to give your comments on whether my understanding is correct.

Thank you in advance…!!

Hi @mneranjan31,

These are great questions and your observation is correct that there’s an inverse relationship between depth and retained sequences (after a point that is).

The goal isn’t to necessarily maximized the depth, or to maximize the sequences retained. You also don’t want to maximize your sampling depth, as that will always leave you with a single sample. Rather, you should evaluate if all of the samples you need are retained.

This might mean you have a hard choice to make, where you could leave a sample in your analysis, at the cost of significant read depth for your other samples. Or you might decide to maximize your read depth up to a sample, but obviously at the cost of any samples beneath that depth.

Generally speaking, any depth you do pick, should be the depth of whatever the last sample you wish to include is. That way you maximize the read depth for everything else (while still keeping that sample you wanted to include). Where that exact point is doesn’t have a hard and fast rule, and if your data forces you to make too many hard choices, a good option is to run it more than one way to see if your conclusions are robust to those choices.


Hi dear Evan Bolyen,
Thank you very much for your comprehensive explanation regarding the sampling depth problem. It makes my understanding more clear to proceed with further practices.

Thanks in advance…!!


This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.