[R-sig-eco] CSS normalization of amplicon data in ordination?

[microbiomics]

Hi all,


Since this is my first post in this list: Hello to everyone reading!

In order to use ecological count data in ordination methods like 
Principal Component Analysis (PCA) or Redundancy Analysis (RDA), they 
have to be transformed. Ecologists [1,2] proposed a number of useful 
transformations for count data, such as chi-square or Hellinger 
transformation, which are also available in the decostand function of 
the vegan package.

With the advent of high-throughput sequencing, ordination methods have 
been applied to metagenomic sequence counts (namely of 16S rDNA 
amplicons). I have been using Hellinger transformation with such data in 
combination with RDA for quite a while with satisfactory results.

In the past years, new normalization techniques specifically suited for 
metagenomic count data, namely cumulative-sum scaling (CSS) [3] were 
developed. The authors claim that their CSS method is superior to 
traditional total-sum scaling (TSS). TSS is also the first step in 
Hellinger transformation (followed by taking the square root).

I am thus wondering if I should switch to CSS transformation of 
amplicon counts for use in RDA - probably in combination with 
square-root transformation to lessen the impact of highly abundant taxa.

Any experience with this matter or suggestions from your side?

Thanks in advance for sharing your thoughts with me!






References:

1. Legendre P, Gallagher ED. 2001. Ecologically meaningful 
transformations for ordination of species data. Oecologia 129:271-280.

2. Borcard D, Gillet F, Legendre P. 2011. Numerical ecology with R. 
Springer, New York.

3. Paulson JN, Stine OC, Bravo HC, Pop M. 2013. Differential abundance 
analysis for microbial marker-gene surveys. Nat Methods 10:1200-1202.