Questions on PCA

[George Tzotzos]

I'm attaching the output of the PCA of an ensemble of proteins. My questions are:

1. What are the units of the y-axis
2. Whether, and if so how, can I determine the degree of correlation between say PC1 and PC2 (these appear to be highly correlated in the figure)

Another question has to do with the plot(pca) function. I understand from the documentation that it is a wrapper function plot.pca.score and plot.pca.scree. Is there a way to separate the two.

Thank you in advance for any suggestions

[George Tzotzos]

Apologies for omitting to attach the output of PCA mentioned earlier. Here it is

[Xinqiu Yao]

Hi,

1. There is no unit (Each bar is a normalized contribution of the corresponding residue to the movement described by the eigenvector or PC).
2. I don't understand this question: What correlation exactly were you going to calculate? (If you just want to characterize the correlated pattern of the two lines, why not use Pearson correlation provided by R i.e., cor()?)
3. Of course. Just call the two functions separately. (See help (plot.pca.score) etc. for more details).

[George Tzotzos]

Thank you for the prompt reply. Concerning the question regarding correlation between PC1, PC2, PC3: In the figure it is clear that the fluctuations overlap to a large extent. The differences between the PCs are more in the magnitude of fluctuations rather than in the mobilities of different parts of the backbone in the MSA. I've applied the rmsip() function to obtain the overlap between nma modes and pca. I understand that in function "rmsip"(modes.a, modes.b, subset=10, row.name="a", col.name="b", ...) modes.a and modes.b can be an object of class "pca". I tried various permutations of syntax trying, what I thought to be, modes.a and b of class "pca" to no avail. Any suggestions of possible syntax would be much appreciated