Can PCA be used to draw a phylogenetic tree?

[Cheng]

I combined dcd data of multiple mutants into one, then run the PCA.

> plot(pc, col=bwr.colors(nrow(xyz)) )

There are 5454 dots, every 303 dots is from a mutant. Can I use the PCA information to draw a phylogenetic tree? e.g.

Because there are too many dots, I think the PCA plot could not capture their similarity information quite well.

Thank you!

[Xinqiu Yao]

Hi Cheng,

What do you want to do a phylogenetic analysis over? Are "A", "B", "C", etc. individual conformations or conformational ensembles from different mutant/WT simulations?

[Cheng]

Hi Xin-Qiu, in my case, I would like "A", "B", "C" to be every 303 dots, as they belong to the same mutant. A mutant of 303 frames includes 3 repeats x 101 trajectories.

[Cheng]

I replotted it, with each colour representing each mutant. So 303 dots * 18 mutants = 5454 dots. So can I analyse the similarities (e.g. a phylogenetic tree) among the mutants?

[Barry Grant]

Just do clustering in the PC space?

Note: This has nothing to do with phylogeny...

[Cheng]

I already assign a unique colour for each mutant, as shown in the last plot. Is that as you said for "clustering in the PC space"?

I agree the scattering of different colours could already provide some information about the similarity among the mutants.

[Barry Grant]

I mean cluster on the pc$z component of the pca output. Read some of the online vignettes to get a feel for this.

Using standard R things like dist(), hclust() with Ward.D2 method and plot() to get your tree like dendrogram if that is what you really ;-)

[Cheng]

yes, plot(pc$z[,c(1,2)]) is same as the first plot in plot(pc).

I will read the standard R things, thank you!