Re: the difference of PCA and EDA

From: Hannes Loeffler (Hannes.Loeffler_at_stfc.ac.uk)
Date: Wed Dec 10 2014 - 05:59:37 CST

On Wed, 10 Dec 2014 13:36:43 +0200
Fotis Baltoumas <fbaltoumas_at_biol.uoa.gr> wrote:

> Hello,
>
> There isn't much of a difference, actually. An Essential Dynamics
> Analysis is, essentially (pun intended), PCA directed at seperating
> the configurational space into two subspaces, a physically constrained
> subspace describing local fluctuation and an “essential” subspace
> with a few degrees of freedom in which anharmonic motion occurs that
> comprises most of the positional fluctuations. Since the latter are
> considered to be "essential" for protein behavior and function, this
> PCA is termed "EDA". Whether you perform EDA on an MD trajectory or
> PCA on an NMR ensemble, basically you're doing the same kind of
> calculations.
>
> You can find more on the subject by reading Berendsen's original EDA
> paper:
> http://homepage.univie.ac.at/mario.barbatti/papers/method/essential_dynamics_1993.pdf

This paper describes how to use PCA to separate modes into "essential"
and "non-essential" ones (that's what PCA was invented for), as you say
above, and how this can be used to reduce the dimensionality of phase
space in an MD simulation. The authors dubbed this method "essential
dynamics".

EDA, on the other hand, is something I have not come across yet but
is simply the renaming of an old hat. According to Wikipedia
(https://en.wikipedia.org/wiki/Principal_component_analysis) PCA goes
by several other names...

I personally find this name rather redundant.

Cheers,
Hannes.

This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:23:06 CST