Interpretation of quality plots

[Miri_J]

Hello, this is my first time using Qiime2. After looking at the different interactive quality plots made by other users, I am unsure whether mine looks correct or needs denoising.

This is what my quality plot looks like.

[colinvwood (Colin Wood)]

Hi @Miri_J,

Welcome to the forum! :qiime2:

These quality plots are unusual. Could you give us more information about the sequencing technology used, and/or the type of the data that you imported?

Thanks!

[Miri_J]

I forgot to include, this is what my fastq files look like

@VH01105:5:AAAMVF3HV:1:1101:19027:1000 1:N:0:CATCAGGCGT+CGTGTTACCG
GATGGAAGCTGTTACTGTGCTCACCAAGGAATGAGCATCTGAGCCAGTGACTGAGGAGGTGGCACCTGACCTGGCTTCAGATGTACATGGGTAGGAACTGGCTACTCACATTAGATCAGGAAGACAAACTTCTCATTGGACCTTTTGTCTT
+
CCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCC;CCCCCCCCCC;CCCCCCCCCCCCCCCCCCCCCCC-CCCCCCCCCCCCCCCCCCCCCCC
@VH01105:5:AAAMVF3HV:1:1101:21450:1000 1:N:0:CATCAGGCGT+CGTGTTACCG
GAACAGCACCATCGAACGCAATGCCCACTCCGGATCCCGCGGTCCACCGGGTGCAGGCCCAGCGCGTGCTTGTGCGGGACGGCGAAGAGCACGCCCGCGCTCGTGTCGACGTCGATCTCCCCCAGGATCCGTCCCAGGTTGGCGGCCTTGG
+
CCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCC--CCCCCCCCCCCCCCC;CCCCCCCCCCCCCCCCCCCCCCCCCCCCCCC-CCCCCCCCCCCCCC;CC
@VH01105:5:AAAMVF3HV:1:1101:22776:1000 1:N:0:CATCAGGCGT+CGTGTTACCG
ATATTGATTCTTCCTATCCATGAGCATGGAATGTTCTTCCATTTGTTTGTGTCTTCTTTTATTTCATTGAGCAGGGGTTTGTAG
+

[Miri_J]

Thanks for the prompt reply!

This is metagenomic data based on Illumina DNA reads. There were 5 samples (s1-s5) paired and all prior steps produced 0 errors.

[colinvwood (Colin Wood)]

Hey @Miri_J,

Thanks for the clarification. Do you know of any prior quality-filtering steps that may have been performed on the reads? I ask because (as you have probably seen from other sequence quality plots) there is usually significant variation in the scores.

[Miri_J]

I believe that the sequencing center our lab uses performs trimming and removes adapter sequences as part of a quality check.

[colinvwood (Colin Wood)]

@Miri_J,

Thanks for the info. Trimming and removing adapters shouldn't have affected the quality scores. I encourage you to reach out to your sequencing center (or anyone else who may have handled the sequences before you) and ask for a description of all the steps that have been performed on them.

Once you have full confidence in the accuracy of these quality scores (which may be the case already), you can move forward with your analysis. In this case, no, you probably wouldn't need to denoise the sequences. Instead you should see the options available in the qiime vsearch command. See the flowchart here to get a rough idea of the workflow.

As a final disclaimer, it would probably be best as a first step to pin point the reason for these very consistent quality scores.

Thanks, and let us know if you have any other questions!

[llenzi (Luca)]

Hi @Miri_J,
These sort of quality scores could be normal, the latest Illumina protocols (eg on NovaSeq runs) produce binned quality scores, very much like the quility scores you have.

The suggested protocol is still to apply a denoising step. however you can not use dada2 for that, because the binning violates its basic assumptions, hence you need to use deblur.
Hope it helps
Luca

[Johanndb (Johann De Beer)]

Dear @llenzi

If I may interject. Why would one not be able to use Dada2? What assumption does this violate? As the field of gut microbiomes expands, it is generally accepted that we should move towards ASVs. However, I don't think deblur clusters according to ASVs.

The reason I am asking is that I received reads from a NovaSeq platform (16s rRNA V4-V5) and have more variable quality scores (see below).

Thank you for your time!

[llenzi (Luca)]

Hi @Johanndb,

you are correct, the general way to do is to move towards ASVs, deblur does indeed ASVs.

On dada2, the model it uses expect the raw quality scores, the NovaSeq platfoms output binned quality scores, hence not the raw data. You can see this dada2 thread for some info regarding this:
https://github.com/benjjneb/dada2/issues/791

On your quality scores, it seems the quality scores in your run are lower than the plot shown at the begin of this thread, however I think deblur will work fine to denise these and produce ASVs.

Hope it helps
Luca