I conducted a Distance-based redundancy analysis (similar to RDA or PCoA) in the R package vegan (Oksanen et al. 2016) on both on both a Bray-Curtis and on a weighted, rooted, Unifrac matrix of 16S bacterial community data. I used the ordistep function to perform forward model selection, determining if my environmental variables explained the bacterial endophyte communities. Models were then reduced by removing non-significant terms sequentially.
My results for the this same analysis using Unifrac and Bray Curtis were different. For Bray Curtis, salinity, water level and tree density contributed to the model. For Unifrac, only tree density contributed.
I know Unifrac takes into account phylogeny, so is this saying only tree density influences communities via phylogenetic relatedness of member species?
Yes. And present both results together with the limitations of the analysis. Or choose the one you like best! :)
Thanks Andrew Paul McKenzie Pegman ! The Unifrac explains less than 1% of the variation on the model, whereas Bray Curtis Explains 16%, which is still low but perhaps that is a better choice. Though it is still interesting to note the difference with the unifrac.
Also, any suggestions on how to rephrase "tree density influences communities via phylogenetic relatedness of member species" so it is more clear?
That title is very good. I think you should use it 'as is' :)
- Badges
- Science topic
- Similar topics
- Biological Science
- Microbiology
- Microorganisms
- Bacteria