Study design Qeustion

Hello Everyone! A big thanks for the Qiime Team and others who are contributing to the forum.

I have 16s data (feces) from an inbred panel of mice - I have 6 animals per line (3 males and 3 females) for about 25 lines. I am trying to answer the following questions

  1. Is the microbiota different between the lines?
    Ans : Beta diversity group significance - PERMANOVA results.

  2. What are the major OTU’s that make them different?
    Ans: ANCOM - the highest W values OTU’s

  3. Are the animals within the same line have similar microbiota? (to make sure I have a homogenous line - feces was collected during various experiments and time through a period of 2 years )
    problem: sample size of six
    Ans: ???

  4. if I know the disease outcomes for these lines, can I predict which OTU’s contribute to resistance?
    Ans: machine learning classifiers - which estimator to select?

  5. If I am using the OTU’s to run some QTLs to see if there are genes associated with the microbiome? which level is appropriate for 16s data
    Ans: Order, family, or genus?

Please add comments or additions to any of these steps or any additional steps which I should do?

Thanks in advance!

Hi @Aravindh_Nagarajan,

This is a really interesting question! I think you’re starting in a good place.

You might try a permdisp and looking at the boxplots from beta group significance, which are traditional ways to look for differences in group dispersion.

I might take the within-line groups and look at the within-group distances to compare similarity. I think I’d use a permuative t-test of the within group distances. Its somewhat controversial as an approach, and would need to be implemented outside of QIIME.

I think in, both cases, its essential that you consider potential confounding in the line, such as cage effects.

The q2-sample-classifer plugin might be helpful here. There’s a tutorial to get you started.

Id recommend starting at the ASV level. I would also be really careful with this analysis because there’s a history of cage, facility, and batch effects in microbiome rodent studies that are easy to over interpret. (I say this as someone who is guilty of it, herself). If you haven’t read it, I found How Informative Is the Mouse for Human Gut Microbiota Research? to be a really good review of this topic.


1 Like

Hi @jwdebelius,
Thank you so much, sorry for the late reply. I had issues running some of the codes.

  1. Different lines went into the same experiments over a period of 3 years, mice were individually housed at least a week before feces was collected
  2. I have filtered the table for ASV’s that less than 2000 and didn’t appear in two samples
    qiime feature-table filter-features
    –i-table MREB_table.qza
    –p-min-samples 2
    –p-min-frequency 2000
    –o-filtered-table MREB_table_min2samples_2kfeatures.qza
  3. rarefaction depth - I picked 7200 based on the feature table

    Alpha diversity

Beta diversity - PERMANOVA

Beta diversity - PERMDISP


  1. Based on the PERMDISP values, does it mean that the samples are not homogenous for any of my metadata groups?
  2. Is there a way to add experiment no to the line and check homogeneity?
  3. ANCOM with line, line+sex picks up almost 50ish OTU’s with the W values in the 60’s, if my sample is not homogenous is it all false?

Thanks for the help!

Hi @Aravindh_Nagarajan,

This question seems beyond the scope of the forum: you’re asking for help with interpretation for your specific study which may border on criteria for authorship in some circles. I’d recommend you find a collaborator - maybe a biostatistican or bioinformatician at your university - who can help interpet your specific data.