> > On Wed, Apr 20, 2011 at 1:06 PM, Kamali Sripathi <ksripath.umich.edu>
> > wrote:
> > you no longer have a function (i.e., -181 degrees and 179 degrees are the
> > same angle, yet both
> > are within your range!)
> >
> > This is actually what I often see, so sometimes it's difficult for me to
> tell which angle (i.e, -181º or 179º) is the real dihedral angle.
Both are the "real" dihedral angle and it is always a question of
interpretation. Moreover, if the torsion values actually populate values
across the whole range from 0->360, a straight coordinate average will not
work and there is no way to range shift to avoid discontinuities in a
non-circular/non-radial plot.
In ptraj, you can add in at the end of the file (when calculating
dihedrals, distances, etc):
analyze statistics all
...and this will show distributions and averages. Here I calculated a DNA
gamma angle (noting the extra "tag gamma" which provides extra summary
information about these angles).
dihedral g0 out dihedrals/g0.dat :1.O5' :1.C5' :1.C4' :1.C3' type gamma
The results are:
STATISTICS g0
AVERAGE: 53.0966 (47.7803 stddev)
INITIAL: 42.9687
FINAL: 60.2678
g+ a+ t a- g- c
-------------------------------------------------------
%occupied | 89.7 | 0.4 | 5.3 | 0.1 | 4.5 | 0.0 |
average | 59.8 | 104.0 | 177.1 | -111.0 | -63.7 | -20.2 |
stddev | 10.2 | 19.4 | 10.6 | 23.0 | 10.2 | 8.5 |
-------------------------------------------------------
GAMMA canonical minor minor
O5'-C5'-C4'-C3', SNB range is 20-80 (g+)
To get the correct average when the angle populates a wide range, you need
to use circular averaging (and this is what ptraj does with the analyze
routines), i.e.
arctan2( summation( sin(a) ) / n + summation( cos(a) ) / n )
(search "circular averaging" in google for more information).
As mentioned previously on the list, I think by Simmerling, for "plotting"
we typically just show the dots/points (not the full line) since then you
do not see these large jumps in the values when changing from -179 to
+181.
In the analyze.c code is special recognition for certain angles that have
defined ranges so you know what the offsets should be (i.e. if you know
the angle tends to populate +60 you can range shift as you are doing the
averaging / progression to avoid the discontinuities), however in general
there is not a great solution if the angle actually populates fully
through 0-360 except not to show the line or to use radial plots.
--tec3
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Apr 27 2011 - 09:30:04 PDT
