Insights of PCA calcilation

[Former user]

Hello sir, here is what i understood to calculate PCA from MD trajectories: (i) first construct the co-variance matrix of calpha atoms of the protein (ii) diagonalize and find the eigen value and corresponding eigen functions and eigen functions with higher variance will give us the principal components.

i have no problems with results.

My question is while constructing the positional fluctuation covariance matrix, are you considering the "center of mass of the atoms" or the "xyz cartesian coordinates of the atoms" for each frame?

thanks in advance kajwal

[Lars Skjærven]

See help(pca.xyz) and the mass argument. By default, no mass weighting is performed.

[Lars Skjærven]

certainly not a bug

[Kajwal Kumar Patra]

i have already gone through help(pca.xyz) but didn't find any solution. i just want to know what are the inputs(xyz coordinates or the cente of mass of the atoms) for constructing the co-variance matrix which will be diagonalize for PCA analysis.

i have provided dcd files. so i think xyz coordinates of each atom of each frames are taken as inputs variable for the covariance matrix construction.

just needed a satisfactory explanation.

[Xinqiu Yao]

Hi,

It just takes whatever coordinates you input to calculate the covariance matrix, i.e. the xyz of each atom of the dcd in your example.

[Barry Grant]

answered