Re: [AMBER] Cluster analysis: the result was strange

From: Nada Afiva <nadaafiva.gmail.com>
Date: Fri, 26 Jun 2020 04:52:23 +0700

Okay, thanks so much, Christina.

Best regards,
Nadaafiva


Pada tanggal Kam, 25 Jun 2020 pukul 20.22 Christina Bergonzo <
cbergonzo.gmail.com> menulis:

> Hi,
>
> The manual is the first link on this page: http://ambermd.org/GetAmber.php
> There is a thorough section on clustering - look for the Analysis Commands
> in Cpptraj.
>
> To check the clustering metrics, you can look for the following (grep) in
> the log file (assuming you have printed the cpptraj output to a log file):
> DBI is Davies-Bouldon index
> pSF is pseudo F
> Number_of_noise_frames will give you how much of the trajectory frames is
> classified as noise
>
> What does the rms plot look like? Does it show a lot of motion? If the
> system is restrained, it might be reasonable that you got a single cluster.
> Alternatively, you can try using kmeans (or another distance-based
> clustering algorithm) which will force a certain number of clusters.
> I'd suggest starting with 3 or 5, saving the PairDistance file, and then
> re-running the clustering with clusters=6,7,8,9,10 and the loadpairdist
> command to see what the DBI/pSF metrics are.
>
> Good luck!
> -Christina
>
> On Thu, Jun 25, 2020 at 4:32 AM Nada Afiva <nadaafiva.gmail.com> wrote:
>
> > Thanks so much Sir.
> > What was the cluster population? The cluster population was 1.
> > How many clusters did you obtain? I obtained 1 cluster only.
> > Did the cluster pop. vs time plots look OK for them? Yes it showed one
> > horizontal straight line.
> > Did you look at how many frames were clustered vs how many noise frames
> you
> > obtained. There are 20000 frames clustered and no noise frame.
> > Did you look at any clustering metrics like DBI, pseudo F, etc? I do not
> so
> > understand it.
> > Did you try to optimize your input parameters for the DBScan algorithm
> > using K-dist plots as described in the manual? Not yes Sir, where is
> > the manual.
> >
> > Thanks.
> >
> > nadaafiva
> >
> > Pada tanggal Rab, 24 Jun 2020 pukul 19.16 Daniel Roe <
> > daniel.r.roe.gmail.com>
> > menulis:
> >
> > > Hi,
> > >
> > > We need more information. What was the cluster population? Were the
> > > rest of your clustering results reasonable? How many clusters did you
> > > obtain? Did the cluster pop. vs time plots look OK for them? Did you
> > > look at how many frames were clustered vs how many noise frames you
> > > obtained? Did you look at any clustering metrics like DBI, pseudo F,
> > > etc? Did you try to optimize your input parameters for the DBScan
> > > algorithm using K-dist plots as described in the manual?
> > >
> > > It would also help if you stated the version of cpptraj you are using.
> > >
> > > -Dan
> > >
> > > On Sat, Jun 20, 2020 at 4:58 AM Nada Afiva <nadaafiva.gmail.com>
> wrote:
> > > >
> > > > Dear Amber,
> > > > I performed a cluster analysis using the input below.
> > > > However, I got a strange result in which the file
> > > > "cluster-cpopvtime-x1d1.agr"
> > > > show one straight line only.
> > > > Hope you can help me.
> > > > Thank you in advance.
> > > >
> > > > Best,
> > > > nadaafiva
> > > >
> > > >
> > > > Here is my input file
> > > > # Cluster analysis with cpptraj.
> > > >
> > > > parm x1d1.prmtop
> > > > trajin nowat.prod.nc
> > > >
> > > > cluster C0 \
> > > > dbscan minpoints 25 epsilon 0.9 sievetoframe \
> > > > rms :1-1079.CA \
> > > > sieve 10 \
> > > > out cluster-x1d1.dat \
> > > > summary cluster-summary-x1d1.dat \
> > > > info cluster-info-x1d1.dat \
> > > > cpopvtime cluster-cpopvtime-x1d1.agr normframe \
> > > > repout rep-cluster-x1d1 repfmt pdb \
> > > > singlerepout cluster-x1d1.nc singlerepfmt netcdf \
> > > > avgout cluster-avg-x1d1 avgfmt restart
> > > > _______________________________________________
> > > > AMBER mailing list
> > > > AMBER.ambermd.org
> > > > http://lists.ambermd.org/mailman/listinfo/amber
> > >
> > > _______________________________________________
> > > AMBER mailing list
> > > AMBER.ambermd.org
> > > http://lists.ambermd.org/mailman/listinfo/amber
> > >
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
>
>
> --
> -----------------------------------------------------------------
> Christina Bergonzo
> Research Chemist
> Biomolecular Measurement Division, MML, NIST
> -----------------------------------------------------------------
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Jun 25 2020 - 15:00:03 PDT
Custom Search