Is my truncation parameter exploration sufficient, and is my interpretation correct?

[Rakaya]

2_trimmed.qzv (368.8 KB)

Hi everyone,

I’m working with 16S rRNA V3–V4 data in QIIME 2/DADA2 and tried to choose truncation parameters by systematically exploring forward and reverse truncation lengths, similar to the approach discussed in this thread Help/advice on truncation parameters - #9 by Deyan_Donchev .

I’ve attached the quality score plots + retention heatmaps for three representative samples.

From these results, my interpretation is that:

• very short reverse truncation lengths reduce merging and retained reads, as expected,
• beyond approximately 180 bp for the reverse reads, the retained non-chimeric reads become fairly stable across a broad range of truncation combinations,
• instead of a single sharp optimum, the data appear to have a plateau where moderate changes in truncation lengths produce very similar results.

My questions are:

1. Does this look like a sufficiently broad exploration to confidently choose a truncation “sweet spot,” or would you recommend exploring the upper-right region (longer forward and/or reverse truncation lengths) further?
2. Is my interpretation correct that this plateau suggests my dataset is relatively insensitive to moderate changes in truncation lengths, and that this is generally a good sign of robustness rather than a limitation?
3. If my interpretation is incorrect or incomplete, could you explain how you would reason through these heatmaps when selecting the final truncation parameters? I’m trying to understand the principles behind the decision.

I’d really appreciate any feedback. I’m relatively new to this aspect of DADA2 and would like to make sure I understand the reasoning behind parameter selection.

Thank you!

[colinbrislawn (Colin J Brislawn)]

Hello Rakaya,

I really appreciate a visual result like this! I can tell you have thought a lot about this.

@Rakaya

From these results, my interpretation is that:

• very short reverse truncation lengths reduce merging and retained reads, as expected,
• beyond approximately 180 bp for the reverse reads, the retained non-chimeric reads become fairly stable across a broad range of truncation combinations,
• instead of a single sharp optimum, the data appear to have a plateau where moderate changes in truncation lengths produce very similar results.

I concur! And this is pretty much expected for Illumina data.

Note that the plots show "% non-chimeric read per sample", which is true, but this implies that chimeric levels change with trimming, which they do not. I suspect with fill="% merged" this would be clearer!

Does this look like a sufficiently broad exploration

Sure, and this is a great check to run when working with new primers or a new sequencing technology! Consider a test that you can apply to all your samples, not just three. Consider using GitHub - Zymo-Research/figaro: An efficient and objective tool for optimizing microbiome rRNA gene trimming parameters · GitHub, if you have not found it already.

this plateau suggests my dataset is relatively insensitive to moderate changes in truncation lengths

Yes, and this is normal for Illumina runs.

and that this is generally a good sign of robustness rather than a limitation?

Uh, neutral. It's just how DADA2 works with Illumina data. I find I can use the same DADA2 trimming settings for a primer set for years, so these settings are robust as long as the primers stay the same.

This is a great example of how the trimming setting work for DADA2!
Thank you for adding this to the forums.

Good job!