differences of PCA analysis with PLUMED and Gromacs (gmx covar and gmx anaeig)

[qingh...@gmail.com]

Hello,

I am doing PCA analysis with both PLUMED and Gromacs, but I got different results for the same trajectory.

With PLUMED, I used:

 domain: GROUP NDX_FILE=index.ndx NDX_GROUP=Bdomain
 PCA METRIC=OPTIMAL ATOMS=domain STRIDE=1 NLOW_DIM=2 OFILE=pca-comp.pdb

and I got the first two vectors in pca-comp.pdb.

With Gromacs, I used:

gmx covar -f prot.xtc -s prot.pdb -n index.ndx

and I got eigenval.xvg, average.pdb, eigenvec.trr, covar.log

Then I used " gmx trjconv -f eigenvec.trr -s average.pdb -o eigenvec.pdb " to view all the vectors in the pdb file,

and I found the differences.

For the first vector:

pca-comp.pdb got from PLUMED (I corrected the residue name and ID):

ATOM      1  N   LYS      1     0.001  -0.697   0.577  1.00  1.00
ATOM      5  CA  LYS     1    -0.055  -0.655   0.557  1.00  1.00
ATOM     23  C   LYS     1    -0.032  -0.606   0.509  1.00  1.00

eigenvec.pdb from Gromacs:

ATOM      1  N   LYS     1       0.090   0.749  -0.555  1.00  0.00           N
ATOM      5  CA  LYS     1       0.142   0.704  -0.532  1.00  0.00           C
ATOM     23  C   LYS     1       0.111   0.653  -0.487  1.00  0.00           C

Then how should I understand the difference?

(As I am planning to bias the first two or three principal components, I have to make sure the vectors are correct.)

Thanks a lot!

All the best,

Qinghua