Re: [AMBER] cpptraj: dihcovar and projection need trajin to work properly?

From: Daniel Roe <daniel.r.roe.gmail.com>
Date: Wed, 1 Mar 2017 13:01:44 -0500

On Wed, Mar 1, 2017 at 12:54 PM, Thomas Fox <thomas_fox.gmx.net> wrote:
> The problem is, that when the number of frames in the input trajectory and
> the number of data points in the data set dont match, cpptraj still produces
> seemingly valid results (and there are no warnings, one has to carefully
> check the size info after "DATASETS:")

Agreed - this is why I would likely remove the 'dihcovar' etc
functionality from 'matrix' entirely. It will break backwards
compatibility, but I think as long as there is a clear message as to
how to modify any existing script that shouldn't be too bad, and will
prevent bad results of the type you describe.

In the short term I could put a warning into 'matrix' that triggers if
the input data sets appear to already be populated.

-Dan

> So when the number of input datapoints Np is larger than the number
> trajectory frames Nf, then only Nf result lines are produced. What is in my
> eyes even more critical, when Nf > Np, I get Nf result lines with numbers
> which look ok, but are obviously drawn out of thin air - at least I havent
> recognized a pattern where they come from...
>
> Thanks for looking into that,
> Th.
>
> Gesendet: Mittwoch, 01. März 2017 um 17:26 Uhr
> Von: "Daniel Roe" <daniel.r.roe.gmail.com>
> An: "AMBER Mailing List" <amber.ambermd.org>
> Betreff: Re: [AMBER] cpptraj: dihcovar and projection need trajin to work
> properly?
> Hi,
> This behavior occurs because of the way cpptraj (like ptraj before it)
> is set up to handle 'Actions'. By definition, an 'Action' is something
> that operates on input trajectory frames, which is why the 'matrix'
> Action requires input trajectory frames. In reality the construction
> of matrices from derived data (like dihedral covariance, IRED, etc)
> should be an 'Analysis', but is in the 'matrix' Action for historical
> reasons.
> I should probably create an Analysis called 'buildmatrix' or something
> for creating matrices from derived data in this manner. I'll open up
> an issue on GitHub and try to get to it ASAP.
> Thanks for the report! Let me know if anything is unclear or if you
> have more questions.
> -Dan
> On Tue, Feb 28, 2017 at 10:33 AM, Thomas Fox <thomas_fox.gmx.net> wrote:
> >
> > Hi,
> >
> > I found a cpptraj behavior that I think is at least unintuitive:
> >
> > From a previous cpptraj run, I have a data set of dihedrals which I would
> > like to further process in a separate cpptraj run.
> > So I use the "readdata" and "dataset mode torsion" commands followed by
> > "matrix dihcovar dihedrals".
> > However, unless I have read in a structure with trajin before (a single
> > frame is enough, no matter how large my data set is),
> > the size of the resulting matrix is 0.
> >
> > I find this weird, as I dont need the parm or trajin info - I exclusively
> > work on the data set I got from readdata.
> > The same is true for the projection command...
> >
> > Ah, and this occurs both with Amber16 cpptraj and with the (current?)
> > cpptraj (it says version V17.00b) I got from github.
> >
> > I have attached my cpptraj input file as well as the dihedral data file
> and
> > the output files for the two cases.
> >
> > Best,
> > Th.
> >
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > [1]http://lists.ambermd.org/mailman/listinfo/amber
> >
> --
> -------------------------
> Daniel R. Roe
> Laboratory of Computational Biology
> National Institutes of Health, NHLBI
> 5635 Fishers Ln, Rm T900
> Rockville MD, 20852
> [2]https://www.lobos.nih.gov/lcb
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> [3]http://lists.ambermd.org/mailman/listinfo/amber
>
> References
>
> 1. http://lists.ambermd.org/mailman/listinfo/amber
> 2. https://www.lobos.nih.gov/lcb
> 3. http://lists.ambermd.org/mailman/listinfo/amber
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber



-- 
-------------------------
Daniel R. Roe
Laboratory of Computational Biology
National Institutes of Health, NHLBI
5635 Fishers Ln, Rm T900
Rockville MD, 20852
https://www.lobos.nih.gov/lcb
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Mar 01 2017 - 10:30:02 PST
Custom Search