QIIME 2 Meta-Analysis of Rhizosphere Microbiome with Mixed V3-V4/V4 Data: Best Practices for Downstream Analysis

[timanix (Timur Yergaliyev)]

@John_Kim_Aligato

We are still unsure if we should be truncating the studies individually with the same parameters when using denoising-single in DADA2 or just truncate them independently of each other?

I would still run Dada2 separately for each study but with the same settings.

@John_Kim_Aligato

We think the single-orientation study was originally paired and then merged? Is there a workaround for this or would we be better off excluding it from our analysis and proceed with just analyzing all of our studies as paired reads and merging them?

You can check it based on the quality plot (check figures 2 and 3 from this post) or by checking the original paper.

@John_Kim_Aligato

If merging stats aren’t looking good for us, how do you suggest we perform downstream analysis from there?

Check your dada2 settings and try to improve the output.

Considering that you decided to work only with V3-V4 region, I would choose between:
Option 1. Work only with paired reads. Then run dada2 for each study separately but with identical settings and merge the outputs.
Option 2. Merge paired reads with VSEARCH (included in qiime2). Then use Deblur instead of dada2 (I think you can even run all the studies together with Deblur).
Option 3. Use only forward reads and trim them at the same position, including datasets with single reads.

I like options 2 and 1. But check if the dataset with long single reads was sequenced with Illumina. If not, check how it can be denoised - if it is incompatible with Dada2 and Deblur, and you decide to keep it, then there will be option 4.

Option 4. Merge paired reads with VSEARCH. Proceed with VSEARCH to OTU clustering (100, 99 or 97%), chimeras removal and dereplication.

All mods who are reading (don't want to spamm with @): Please join the discussion if you have better options in mind. I will leave it queued for a while.

Best,
Timur