It appears that using a sequencing depth of 1000 is appropriate to capture the diversity in my samples. However is that 1000 reads per sample? 1000 sequences per sample? I am writing up my methods section and would like to be as descriptive/accurate as possible. Thanks!
Hi @sabitondo,
This graph looks great!
The units are feature counts per sample. The 'Moving Pictures' tutorial has descriptions which may be helpful to you in your methods write-up (this one is specifically about alpha-rarefaction plotting).
Also, I would recommend that you rerun your alpha-rarefaction step and adjust your max-depth until you start seeing some samples drop off, that way you know you are getting the most out of your analysis. Right now it looks like most (all?) of your samples run to the end of the max-depth of 10000.
Or check out qiime feature-table summarize and use the visualization there to optimize your sampling depth.
I hope this is helpful to you!
-Hannah
Hi Hannah! Thanks so much for your quick reply. I redid my graph this time with a p max depth of 30,000 and got the resulting figure where some samples definitely have dropped off:
Does this change anything for downstream analysis? Because the samples still level off in the same place? Previously when I computed alpha diversity I used:
would you maybe choose a higher --p-sampling-depth for computing core metrics like maybe ~12,000 since that is before the sample that drops off first? Or does it not matter since diversity appears to be similar for all samples at a depth of 1,000 and all the way to 12,000?
Hi @sabitondo, This forum post suggests that because your alpha rarefaction plots are leveling off that your sampling depth is at a good place for capturing diversity.
But you are correct, when calculating core metrics you will want to use a sampling depth of 12,000, at least, to maximize the features included in your study.
-Hannah
