[gmx-users] Principle component analysis cosine content

[sridhar at cdfd.org.in]

Dear All,

I would like suggestions on the following questions in Essential Dynamics
analysis. I have been working on the essential dynamics analysis of a 450
aa protein and its mutant.

To exclude the existance of random diffusion effects, I calculated cosine
content and found it to be very high for the 1st principle component
(~0.9), and very low (< 0.02) for subsequent PCs.
Later to just check I did the same analysis using the starting structure
as reference for g_covar. I found that the cosine content has dropped down
to ~0.2 for the first PC.
My question is why the cosine content is dependent on the fitting
structure used in g_covar?  And, especially, which structure is
recommended to use in Essential Dynamics analysis of a combined
trajectory? If we do have PCs with high cosine content, how should we
interpret the results?


Thankyou so much for your suggessions.

kind regards,

sridhar