I am trying to do PCA from the MD trajectory of a multi-domain protein using GROMACS tools.
I am using gmx covar and gmx anaeig for PCA. My protein has many domains, but I would like to focus on only two particular domains. Thus, while doing gmx covar, for lsqft I am using protein-H as the group and for covariance analysis I am using index for one particular domain. Similarly for gmx anaeig, I am focusing only on that particular domain and writing as output (the filter.pdb) for only that domain.
I am not sure it this is physical and correct. In short my query is can I calculate PCA (covariance matrix) only for a particular domain of a protein? or should I always do PCA for the entire protein?
Hi Angana,
Yes, you just have to select the protein regions that form the domain of interest with the gmx make command and perform the same process that is followed for the entire system.
This is perfectly physical and sound to analyze only a part of a protein. Keep in mind though that your extracted data should preserve variance that you are looking to characterize. For instance, if you intend to study how a part of a protein (e.g., a helix or a loop) moves with respect to the rest of the protein, then by extracting only coordinates related to this particular part would result in a complete loss of the information you are interested in.
Thank you Dmitri Kireev .
I understand, if I have to compare motions of different domains withing the same protein, I should calculate PCA for the entire protein (ie my covariance matrix should be for all protein). However, if I wish to compare motion of the same domain but in different protein (eg wt and mutant comparison), I can calculate PCA for that particular domain.
Please correct me if I am wrong.
- Badges
- Science topic