q2-longitudinal analysis results interpretation and seeking for help

We are a Human Microbiome Research (HMR) laboratory in National Institute of Biomedical Genomics, India. Recently, we have generated data on vaginal microbiome (V3-V4 of 16S rDNA) of 18 Preterm and 20 Term mothers for all the three trimesters. For performing longitudinal analysis, we used q2-longitudinal (Bokulich et al., 2018 mSystems). We have performed Linear Mixed Effect Model (LME) in 3 ways. However, we could not interpret the appropriate models to be used or the results obtained. I am attaching the plots and results that is generated but could not interpret these. Can you please provide us necessary guidance on how to use q2-longitudinal for our data?

Hi, @souvikm, and welcome to the forum! :wave:

Would you mind providing more details about why you are struggling to interpret your results? It’s great that you’ve included so many details in your post! :+1: However, it will be quite a lot of work to sift through all of this information in order to try and figure out how to help you. If you can point to specific examples and provide details about exactly what you are confused about, I think you will have much better luck getting help here on the forum.

4 Likes

Dear Andrew,

We have relative abundance data on Lactobacillus iners for 14 Preterm and 19 Term mothers for 1st, 2nd and 3rd trimesters of pregnancy. We want to find out if there is any significant longitudinal fluctuation or shift in the relative abundance of L. iners in both the groups across the three trimesters. For this, we have used q2-longitudinal and performed 3 types of analysis, not sure which one is providing us correct information.

In Case-I, we took, Response Variable=Birth-Type (preterm or term) (Y-axis) & State Column=Taxon (X-axis). However, from the scatter plot the answer to our question was not evident as each regression line was annotated for each trimesters. (refer to figure 1)

In Case-II, we took, Response Variable=Taxon (Y-axis) & State Column=Trimester (X-axis). In this case, we are not able to interpret what the regression lines are showing. However, is it correct to use Taxon (L iners) as the Response variable? Since, response should be term birth or preterm birth as per our understanding. (refer to figure 2)

These are the confusions, please advice.

Thanks for the additional details, @souvikm! Sorry for the late response; I’ll take a look at this in detail and get back to you ASAP.

Cheers,
Andrew

1 Like

Hi, @souvikm ,

Specifying birth type as the response variable means that you want to check if the birth type responds to/depends upon the relative abundance of Taxon. The resulting scatter plot shows no response for Taxon. Can you see why that’s the case?

Additionally, since you mentioned that you are specifically interested in looking at changes in the relative abundance in relation to trimesters, I don’t think specifying the state column as Taxon is what you want to do. To quote from the q2-longitudinal tutorial:

“States” will most commonly be related to time or an environmental gradient, and for paired analyses (pairwise-distances and pairwise-differences) the sample pairs should typically consist of the same individual subject observed at two different time points. For example, patients in a clinical study whose stool samples are collected before and after receiving treatment.
“States” can also commonly be methodological, in which case sample pairs will usually be the same individual at the same time with two different methods. For example, q2-longitudinal could compare the effects of different collection methods, storage methods, DNA extraction methods, or any bioinformatic processing steps on the feature composition of individual samples.

I hope that helps. Let me know if you have any other specific questions. I think if you take the time to carefully work through at least the first half of the longitudinal tutorial I linked to above, it will clear up a lot of things for you.

2 Likes