Re: [AMBER] # of Coordinates Mismatch in PCA

From: Jason Swails <jason.swails.gmail.com>
Date: Tue, 14 Aug 2018 19:55:05 -0400

 

> On Aug 12, 2018, at 1:04 AM, Robert Molt <rwmolt07.gmail.com> wrote:
>
> Good morning,
>
> I am attempting to perform a principle component analysis but am
> encountering an unexpected error during the projection step using cpptraj:
>
> Error: number selected coords (5865) != number avg coords (10983) in myEvecs
>
> In short, I am confused how a mismatch could occur. My understanding is
> that if my eigenvectors (myEvecs) come directly from the covariance
> matrix of a given trajectory, no mismatch should be possible (if I have
> consistent masks). I must be confused and seek clarification.
>
> I am using cpptraj v. 15. I am following the tutorial specified on
>
> http://www.amber.utah.edu/AMBER-workshop/London-2015/pca/
>
> My cpptraj commands are (in blue), with trivial warning messages listed
> (in red):
>
> trajin full_no_waters.mdcrd
> reference Crystal_Structure_with_Hydrogens.pdb
>
> Warning: PDB line length is short (67 chars, expected 80).
> Warning: PDB Crystal_Structure_with_Hydrogens.pdb: Reading frame 2, got
> 711 atoms, expected 4431.
> Warning: Only using frames 1-1
>
> createcrd full_no_waters.mdcrd
> run
> crdaction full_no_waters.mdcrd rms ref
> Crystal_Structure_with_Hydrogens.pdb :38-318&!.H

Note the mask. :38-318&!.H

This selects all atoms in residues 38-318 not including atoms names H.

> crdaction full_no_waters.mdcrd projection full_no_waters_projection3
> modes myEvecs beg 1 end 3 :38-318&!.H=

Note this mask with a = at the end. The = is a wildcard that omits all atoms that STARTS with H. This includes a lot more atoms that are excluded from the mask you’re doing the projection on.

> Error: number selected coords (5865) != number avg coords (10983) in myEvecs

Based on this, there are (10983-5865)/3=1706 atoms that start with H but are not named H.

Make sure you’re using exactly the same mask in each step.

HTH,
Jason

--
Jason M. Swails
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Aug 14 2018 - 17:00:01 PDT
Custom Search