Hello everyone. I am trying to use the relative abundance OTUs data generated from QIIME2 in mixed linear model association. Before doing so I need to generate a “Microbial Relationship Matrix” of the OTUs.
I have reviewed different research papers and some have used methods suggested by Ross et al 2013 while some are using Bray–Curtis dissimilarities distance matrix. I can post more details regarding the methods that can be adopted but its not necessary as I think users working with microbial relationship matrix can understand what I mean.
I’m not entirely sureI understand your question. Do you want a bray-curtis dissimilarity matrix between samples? That can be generated using beta function in the q2-diversity plugin. Or, do you want a matrix describing the relationship between the features?
0,1,2,3.... are the sample number. I need to carry on this table to Microbial Relationship Matrix.
I have also read another approach which involves the microbial relationship matrix based on the variance–covariance matrix from the log‐transformed and standardized OTU table.
It’d help if you can define to us what you mean by the term:
This can mean many different things. Is there a reference from a paper, or other documentation, that you can share with us? This will help us better understand what type of analysis you are trying to perform. I assume you want a comparison between samples (i.e. beta diversity of microbial communities), like Bray Curtis (dissimilarity matrix), UniFrac,… Something else?
Yes. There are many beta-diveristy metrics to choose from, some of which are available in QIIME 2. I’d recommend going through the tutorials and see if what you need is there. Particularly the diversity analysis overview.
Hey Mike, Thanks for jumping in.
I wanted to explain a bit more on Microbial Relationship Matrix earlier but I was assuming someone helping me will be familiar with it
So for now I am looking for constructing Microbial Relationship Matrix by following Ross et. al 2013 methodology. It constructs a relationship matrix based on the variance–covariance matrix from the log‐transformed and standardized OTU table. Such kind of matrix is used in statistical model in many papers, For reference you can check this paper “Metagenomic Predictions From Microbiome to Complex Health and Environmental Phenotypes in Humans and Cattle”.
I have already constructed a Bray Curtis (dissimilarity matrix) but I need to construct multiple relationship matrixes and decide which one suits the model.
In the future, when referring to an article, please provide a link, it’ll save time.
Sadly, I am unfamiliar with this approach. Perhaps someone else on the forum will be able to help? On the surface this approach appears intended for shotgun metagneomics analyses. But I imagine it can be co-opted for amplicon analyses too.
To clarify your question, are you looking for a method / tool that will calculate the Microbial Relationship Matrix for you, given a feature-table, or other input? If so, you may have to ask the Ross et al. 2013 authors, unless others on the forum are able to help.
Yes, for FeatureTable[Frequency]. See the associated documentation about the types I linked before.
A quick forum search lands me on this, which suggests that you can not currently use FeatureTable[RelativeFrequency] as input into bray_curtis, as outlined here. So, for the moment, this metric still requires FeatureTable[Frequency], i.e. counts.
Yes by table-freq.qza I mean the FeatureTable[RelativeFrequency]. It would be nice to use relative abundancy as input for bray-curtis matrix.So I hope to get this feature in QIIME2 in near future. Thanks for the effort guys.
Perhaps someone else on the forum will be able to help? On the surface this approach appears intended for shotgun metagneomics analyses. But I imagine it can be co-opted for amplicon analyses too.