No bacteroidales in just two samples.

Hi every one.
I'm having a weird data output of two samples.
AD2, AD3 shows relatively small amount of observed features in alpha rarefaction curves,(Fig.1) and ends to different aspects of taxonomy result compared to other samples.
As you can see below, the two samples have no "Bacteroidales Bacteroidaceae" but relatively a lot of "Enterobacteriales Enterovacteriaceae", (Fig.2, 3) and some other features that does not follow the average.

I want to know whether this is an error or not, and how should I interpret it or fix it.

Rawdata:
Fig.1


Fig.2

Fig.3

Fig.4

For more information;
These are microbiome data of mice samples (AD: Atopic model, ACU: treatment, CON: naive).
We used both Greengene and NCBI but got the same aspect of results.
metadata.tsv (1.0 KB)
raw reads:
https://drive.google.com/drive/folders/1A821pUeP-tV1JdeDVcp4ya9uGebc_jeG?usp=sharing

Please, I'm very new to microbiomes and don't know where to ask. Any type of source that I can search through would be helpful! :slight_smile:

Hello Sunjeong,

Welcome to the forums! :qiime2:

Did you included any positive controls on your sequencing run? I ask because you can use your positive controls to verify that your processing pipeline is producing the results you expect from these samples with a known composition, which would give you confidence that these unexpected results are also true.

How many reads are in these two samples, compared to the other samples on this run? If samples do not sequence well, that could explain the low alpha diversity you observed in Figure 1.

Hello Colin,

Thank you so much for taking a look.
We actually ran the same analysis with different set of data and they didn't show any significant outliers like these two. Which made me suppose the pipeline is fine.

AD2 showed 82507 and AD3 showed 102007 sequence counts which rank 4th and 15th among the 18 samples. ( if this is what you meant my reads)
And what are the indicators of bad sequence? I thought too many or too small number of sequence would be the indicator, but it doesn't in my samples.

OK, good to know!

These read depths look great, and samples with lower coverage still look more diverse. This is the indicator I was looking for, and it does not seem to be the issue here.

Maybe these samples are not very diverse :person_shrugging: