mothur

corr.axes and biplots

Hi all,

I’m still relatively new to this and I was wondering how I could go about constructing a biplot from the output of the corr.axes command? I have a a PCoA plot based on thetaYC which results in separation of my samples into separate clusters and I have run the corr.axes command to determine which OTUs are causing the shifts along the 2 axes I have plotted. I see the output of corr.axes gives a table with p values and a positive or negative value for each axes for the OTUs. What do these values mean and how would I incorporate them into a biplot?

Thanks in advance for any help anyone could offer on this.

I’m also fairly new to this but from the Schloss SOP I get the impression that you can overlay your biplot from the OTUs on that of your samples and then get a visualization of which OTUs drive the pattern in separating your samples. So just include your OTU coordinates in the plot of your samples, in what ever program your using to make the plot. Does that make sense?.. Hope I’m not confusing things further :slight_smile:

Thanks for your help with that I appreciate it. I’ll have a go plotting the axes values for the OTUs over those of the samples.

I’m still a bit confused about this line in the SOP:These data can be plotted in what’s known as a biplot where lines radiating from the origin (axis1=0, axis2=0, axis3=0) to the correlation values with each axis are mapped on top of the PCoA or NMDS plots.

Also in the output of corr.axes there is a column labelled ‘length’, what does this mean exactly?

I think the length is the distance from the origin (0,0,0) and that this is the line from the origin to the plotted location of the OTU, the length describing how much the OTU explains in segregating your samples along that direction. I’m not completely sure, but does that make sense considering the plots you’ve made?

Bingo, SandraBA - you’re right.

Hello,
I am plotting spearman correlations in a biplot from corr.axes, and I have a question about length. I have a few, I suppose, highly correlating variables with long lengths (0.86 is the longest), and it actually distorts my plot to put them in those positions. Would it be totally out of line to reduce the length of all these arrows by the same amount (0.5 or something) so that the arrows don’t widen the ordination coordinates so much?