Where did it go wrong and I got this result?

I was doing a 5 sample analysis. I’m pretty sure the result shouldn’t look like this, but I have no idea where it got wrong. I’m posting taxa barplot and denoising stats so you could take a look.
I need to know if this is about computational problem or something went wrong in sample preparation process.
If you need to see any other file, I can post it, too. I need help asap.
(This is the first time I upload something, and I can’t preview the post, so I don’t know how it will look like. If it is bad, please excuse me :))

Thank you for your attention dear Qiimers…

You’re going to have to give some more information about the samples, because without knowing what you expect it to look like, it’s hard to say what went wrong (if anything).

The only obvious thing I can see is that your barplot has only 4 samples, not 5; presumably the last (OK1) didn’t have enough reads to be included. You also seem to have lost a lot (~80%) of your reads in the first filtering step; not sure if that might be a problem.

What were these 5 samples? It looks like OK2 & OK3 are one sample type, and OK4 and OK5 are another. Is that the case, or are these supposed to be 5 similar replicates?


I got pneumothorax last week, so I wasn’t around :mask:

I don’t know the reason I lost most of my reads. Quality scores are actually better than many of the analysis I did. Is there something that I could try to keep my reads?

P.S. I have learned that these samples are from a plant epidermis tissue, but they did not explain whole treatment.

1 Like

I have given a try to denoise with only forward reads, and I used same trunc value as I did with paired end run. I’m uploading taxa barplots and denoising stats again. Seems like this time I kept enough reads because feature counts were all above 10000.

I haven’t got a feedback about this result yet, I just wanted to show what happened when I tried something else. I will share it if this is a result which is expected or not.

Yikes! Glad to have you back, @the_dummy, hope you are doing better!

Given your results (but not having seen a critical piece of info: the demux summarize QZV to look at read quality), it sounds like this is what’s happening:

  1. the reverse reads are relatively low quality and need to be truncated more than you were doing.
  2. When using dada2 denoise-paired you lose many more reads at filtering because the reverse reads are not meeting the filtering criteria and are being lost.
  3. When using denoise-single you save many more reads because the reverse reads are not bogging you down.

You are saving enough reads at the filtering step that I’d say you could probably proceed with these data if you want — but you could probably save even more reads if you wanted to adjust the trimming parameters even more (and perhaps even save the reverse reads? maybe not). Share the demux summary QZV and we can take a look.

1 Like

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