[gmx-users] COsine content of ED analysis

Berk Hess gmx3 at hotmail.com
Tue Sep 7 11:05:12 CEST 2004

>that may be, but even while walking in the first "well" the flexibility
>required to jump to the next wells is already available - it is simply
>embedded/described by the fold that the protein is in. In addition there
>are significant overlaps between ED eigenvectors calculated from clusters
>of crystal structures or NMR ensembles (i.e. real experimental data). So
>perhaps it is not totally useless to use simulation times of less than 10

The chances are very small that for "true" motions the first few
principal components follow:
p_i(t) = A/i cos(pi i t/t_total)
with the correct 1/i amplitude dependence, the period matching
the simulation length and phase zero.
But theoretically it could happen.
An easy way to check is do to ED on the two halfs of the trajectory.
For "true" motions one should find a quarter cosine, a half cosine etc.
But I would bet that in all cases you again find a half cosine, a full 
cosine etc.


Play online games with your friends with MSN Messenger 

More information about the gromacs.org_gmx-users mailing list