Statistical assessment of PCA patterns

Hi all,

The sPLS-DA provides a nice visualization of the differences in metabolomes between samples. Is there a statistical methods to support strong clustering within metaboanalyst? I am thinking of something along the lines of an ANOSIM or PERMANOVA.

Thank you,

Tyler J. Carrier, Ph.D.
Department of Biological Sciences
University of North Carolina at Charlotte
Charlotte, NC 28223

It is certainly possible to apply the clustering assessments to PCA (similar to PCoA employed in ecological data analysis).

Not sure if it is very meaningful or useful to apply this to sPLS-DA. The common practice is cross validation and permutation tests for supervised methods.

Hi Jeff,

I can also use PCA; that is no problem. How would one perform an ANOSIM or PERMANOVA–or something similar–using the online version of MetaboAnalyst? I did not see this option. I say ANOVAs for each metabolite, but not a ‘community-wide’ analysis.


No, this is not available in MetaboAnalyst. Maybe you can try our MicrobiomeAnalyst which supports this practice from the very beginning.

Hi Jeff,

But the data are not related to the microbiome, so how would that work?


The feature is now available in PCA 2D Scores Plot in MetaboAnalyst (Statistical Analysis module). Note PERMANOVA is performed on the currently selected 2 PCs.

Hope the information can add some quantitative reassurance to the visual inspection.

Hi Jeff,

this is fantastic!! thank you for incorporating this into MetaboAnalyst! I’ve spent parts of the last month trying to do this in R, but all of the ‘solutions’ that I found did not like the negative numbers from the normalized table.


This topic was automatically closed after 4 days. New replies are no longer allowed.