I performed Principal Component Analysis (PCA) using GROMACS 5.1.3 with the following command
gmx covar -f filename.xtc -s su2-now.gro -n protein-H.ndx -o eigval.xvg -v eigenvec.trr -ascii covar.dat
A covar.dat file was generated with (3n*3n)/3 rows and 3 coloumns where n is the number of atoms on which PCA has been done (indicated by protein-H.ndx file) . I want to know how I can interpret this data. Does the 1st 3n rows correspond data for 1st mode , 2nd 3n rows correspond data for 2nd mode and so on and 3 coloumns correspond to x y and z ?
No, you have a covariance matrix and nothing more. To obtain the modes, you need to run gmx anaeig. But the output of gmx covar is exactly what it says it is - the covariance matrix.
Hi,
A covariance matrix must be a symmetric matrix and it should have the dimension 3N x 3N having variance as it's diagonal element whereas gmx covar is producing covariance matrix containing dataset having a dimension ((3N*3N)/3) x 3 where N is the number of atoms. Why is the matrix not symmetric ?
The format that GROMACS provides is 3N x N, with (x,y,z) on each line. So it is a 3N x 3N matrix, just organized a bit strangely.
I have attached a raw variance covariance ascii data file named covar-1000.dat . I had 585 atoms . It has 1026675 lines which is (3*585)*585. From this I want to construct 3N x 3N matrix . How should I proceed ? Should I take 1st (3*585) lines as 1st coloumn , next (3*585) lines as 2nd coloumn .... 585 th (3*585) lines as 585 th coloumn in variance-covariance matrix (3N x 3N) ?
In this website if I click on 'get the source code' option it gives a .tar.gz files for each of the version. Where can I get the programs which are running behind the command gmx covar ?
- Badges
- Science topic