I am working with 16S rRNA gene amplicon sequencing data in QIIME 2 (version 2026.4), installed through Conda on an HPC cluster. I have generated beta-diversity distance matrices using Bray–Curtis dissimilarity and Jaccard distance.
The issue is that some sample pairs have Bray–Curtis dissimilarity values of 1 against almost all other samples, including samples collected during the same sampling trip. Some of these samples have approximately 90,000 reads.
However, when I inspect the feature-table visualisations and the exported feature tables at taxonomic levels 4, 5 and 6, I can see ASVs shared between samples. This makes the Bray–Curtis results unexpected, and I would like to understand whether the values are correct or whether there may be an issue with the data or distance-matrix processing.
The corresponding UniFrac distances for these samples are below 1.
I have used CASAVA format before for another project and I do not remember extracting JD and BC indices using code. I also do not have metadata, but only index numbers that I use to merge metadata and the BC indices later into R studio. Can please I be assisted with this?
for reference my code for BC was: qiime diversity beta \
When you are comparing collapsed-to-taxonomy-level features, you are dealing with taxonomy annotations, not ASVs. Different ASVs may share the same taxonomy but still differ in sequence.
Was BC calculated at the ASV or taxonomy level? If at the ASV level, see my explanation above.
I have an additional question: Were all samples in the dataset you are working with processed together, with identical primers/trimming/denoising parameters? Differences in the steps above may lead to different ASVs from the same biological sequences.
I just converted the biom feature table to a tsv, and I see multiple samples that have contrasting OTUs that are not shared. It is really weird as they are all seawater samples. I BLAST a few of these shared and not shared OTUS and they all point me toward Synechococcus (different strains).
What would you recommend as the next step to investigate this? I dont understand why most samples show a more consistent pattern, whereas these few samples appear so distinctly different despite all being seawater samples. Could this reflect genuine differences in the microbial communities, or would you recommend checking for potential issues with the data or processing first, although I ran all samples through QIIME2 at once.
I don't think there are any issues related to the CASAVA format. I have worked with that format for years and have never had similar issues. So, unless the feature table was somehow manipulated or files were processed separately, I would assume that there are simply no overlapping ASVs between some samples with BC of 1. You can check it, just use the original table you used as the input, not collapsed to taxonomy table.
Hey I am so sorry, I was in the process of editing the previous message when you replied.
I just converted the biom feature table to a tsv, and I see multiple samples that have contrasting OTUs that are not shared. It is really weird as they are all seawater samples. I BLAST a few of these shared and not shared OTUS and they all point me toward Synechococcus (different strains).
What would you recommend as the next step to investigate this? I dont understand why most samples show a more consistent pattern, whereas these few samples appear so distinctly different despite all being seawater samples. Could this reflect genuine differences in the microbial communities, or would you recommend checking for potential issues with the data or processing first, although I ran all samples through QIIME2 at once.
Since you processed all the samples together, I tend to think that your samples are just different. I would add as much info as possible to the metadata file describing your samples. If there were any differences between the samples during DNA extraction/amplification steps, add those to the metadata as well. Then create a PCoA plot, visualize it in Q2 View, and check whether any known factors explain the differences between samples. It's also common to have a couple of outliers that differ from the rest of the samples, since many factors can affect environmental samples, and not all of them can be documented or observed.
Hi @Ashoi_Dantra,
I just got a tip that you also should check if barcode sequences were removed from the sequences while demultiplexing. I would suggest the following check list:
Were barcodes removed? If not, remove them first.
Were primers removed with cutadapt, or trimmed with Dada2? It is better to remove them with cutadapt before dada2.