[gmx-users] Problem in PCA of protein ligand system

Tsjerk Wassenaar tsjerkw at gmail.com
Mon Sep 26 08:52:11 CEST 2016

Hi Ashutosh,

To simplify this, let's do PCA of two balls on opposite ends of a stick I'm
rotating. The mean position of both ends is right at the center of
rotation, and the relative positions I can describe with X and Y
coordinates only. Now, the essence of PCA is the question 'which single
direction can I find that explains most of the spread of my points?'. Since
this is pure rotation, any direction is as good as any other, so I just
pick the horizontal line through the center. I then project my trajectory
onto this axis. Surprise: I find that the principal component describes
lengthening and contraction of my stick along the horizontal direction.
What? Let's check the other component, orthogonal to the first. That too
describes lengthening and contraction, but anticorrelated with the
projection onto the first. The thing is, I can't ever describe a
((semi-)rigid) rotation with a single component. The projection needs to go
through the center and come out the other end, looking like a contraction
and expansion. Likewise, the mean structure is halfway, so it's the most
distorted configuration along the axis.

I hope this clarifies your observations.



On Mon, Sep 26, 2016 at 4:32 AM, ashutosh srivastava <ashu4487 at gmail.com>

> Dear all
> I have performed a 200 ns simulation on a protein ligand  (MOL) system in
> gromacs 5.1.2. Is it possible to get low frequency motions of only the
> ligand?
> When I am looking at the filtered trajectory (along PC1) after performing
> PCA on the protein+MOL the small molecule looks distorted. There is an
> aromatic ring in the molecule that collapses and expands during the course
> of trajectory and the methyl groups are all collapsed throughout the
> trajectory.
> Following is the command that I am using
> covar -f "$TRAJ" -s "$TPR" -o
> ck2_go289_eigval_${first_frame}_${interval}.xvg -v
> ck2_go289_eigvec_${first_frame}_${interval}.trr -av
> ck2_go289_eigavg_${first_frame}_${interval}.pdb -l
> ck2_go289_eiglog_${first_frame}_${interval}.log -ascii
> ck2_go289_eigcovar_${first_frame}_${interval}.dat -xpm
> ck2_go289_eigcovar_${first_frame}_${interval}.xpm -xpma
> ck2_go289_eigcovara_${first_frame}_${interval}.xpm -b $first_frame -e
> $interval
> anaeig -v ck2_go289_eigvec_${first_frame}_${interval}.trr -f "$TRAJ" -s
> "$TPR" -eig ck2_go289_eigval_${first_frame}_${interval}.xvg -comp
> ck2_go289_eigcomp1_${first_frame}_${interval}.xvg -rmsf
> ck2_go289_eigrmsf1_${first_frame}_${interval}.xvg -proj
> ck2_go289_eigproj1_${first_frame}_${interval}.xvg -filt
> ck2_go289_eigfilt1_${first_frame}_${interval}.xtc -first 1 -last 1 -b
> $first_frame -e $interval
> I have also tried giving -extr option and the extracting 2000 frames along
> the PC1, the molecule looks distorted even in this.
> I have tried following
> 1) Fitting CA and MOL and covariance analysis on CA and MOL
> 2) Fittting CA only and covariance on MOL only
> 3) Fitting MOL only and covariance on MOL only
> 4) Isolating the MOL trajectory separately and then performing PCA on this
> trajectory.
> 5) Fitting all atoms and covariance analysis on all atoms (protein+MOL).
> Before all this I performed following PBC corrections on the trajectory.
> trjconv -s prod_0-10.tpr -f prod_0-100.xtc -o whole_100ns.xtc -pbc whole
> trjconv -s prod_0-10.tpr -f whole_100ns.xtc -o nojump_100ns.xtc -pbc nojump
> trjconv -s prod_0-10.tpr -f nojump_100ns.xtc -o center_mol_ur_compact.xtc
> -pbc mol -center -ur compact -n index.ndx
> trjconv -s prod_0-10.tpr -f center_mol_ur_compact.xtc -o
> fit_center_mol_ur_compact.xtc -n index.ndx -fit rot+trans
> I have done mass weighted analysis and the result is the same.
> In all the cases the MOL becomes distorted in the filtered trajectory.
> In order to visualize the trajectory I am extracting the 0th frame of the
> trajectory (CA and MOL) from the actual trajectory and then loading
> filtered trajectory on top of that.
> I also tried extracting the 0th frame of filtered trajectory and then
> loading the trajectory on top of that.
> I have looked at the actual trajectory  and the small molecule as well as
> protein is fine in that with no weird motions.
> What am I doing wrong here?
> Kindly let me know if anymore details are required.
> Thank you
> Ashutosh.
> --
> Gromacs Users mailing list
> * Please search the archive at http://www.gromacs.org/
> Support/Mailing_Lists/GMX-Users_List before posting!
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.

Tsjerk A. Wassenaar, Ph.D.

More information about the gromacs.org_gmx-users mailing list