Here is email I exchanged with Chris Greenwell about using Chimera for a material science / molecular dynamics study. Tom --------- Date: Thu, 21 Oct 2004 09:14:55 +0100 Subject: Re: Material visualization with Chimera From: "H. Chris Greenwell" <h.greenwell@ucl.ac.uk> To: Thomas Goddard <goddard@cgl.ucsf.edu> Hi Tom, First of all, Chimera is fantastic! Congratulations! Yes, I noticed the problem with large numbers of atoms. We run systems getting on for 500,000 atoms., but most are less than 100,000 atoms. I still use the Atom Eye software, which is more efficient at using memory (http://164.107.79.177/Archive/Graphics/A/) for large models. However, it is less intuitive to use and doesn't have much of the functionality that we would like. The type of work we do at the moment involves simulating the interlayer region of layered clay type structures with organic molecules, such as long monomers, between the crystalline clay sheets. We do this to gain insight into the interlayer arrangement and dynamics not attainable by experimental means. We have to present data which is both meaningful and interesting to the experimental community. We use the LAMMPS code as our large-scale MD engine, from Sandia. We have our own data analysis tools to generate RDFs, diffusion coefficients and atom density plots. However, in order to make sense of the features, and to spot unusual features that get averaged in the data of these numerical data sets it is necessary to visualize the model through the simulation. As an ex-experimental chemist I have a strong bias to the visual! I have looked at many types of software, and essentially we have two considerations: 1) A visualizer for looking at each frame, and movies of the atom positions. Data analysis as described above. 2) A builder for loading crystalline unit cells, generating organic monomers, translating, rotating and deleting atoms/molecules. Critically, we tend to use periodic boundary conditions in all our simulation so an ability to replicate unit cells and visualize the boundary markers when desired. Functionality to place forcefield atom types on the system is also needed. I have come across only two contenders to do this with: ECCE (http://ecce.emsl.pnl.gov/) looks promising and is free to academia, however, as with many freeware builders/visualizers, it is only for bio-systems and has no concept of periodic unit cells. I also am unaware of the atom number capability. Materials Studio/Cerius 2. These are the commercial options from Accelrys and work very well. Unfortunately, these are also incapable of building models greater then approx. 30,000 atoms for materials science. Materials Studio has a bug, which stops it doing this, and Cerius 2 (a) gets very slow, and (b) has a purely numeric atom numbering, which has a 5 digit capacity and relies on the molecules being made up of residues. We are starting to use the latest LAMMPS2003 software from Sandia, and may well look at steering it in the future, so any developments in Chimera to aid us in looking at materials would be much appreciated. Many thanks Chris On Thursday, October 21, 2004, at 12:02 AM, Thomas Goddard wrote:
Hi Chris,
Dave Konerding told me you tried using Chimera to visualize materials with large numbers of atoms. Unfortunately Chimera uses alot of memory per atom (~2 Kbytes) and gets pretty slow for systems of 100,000 atoms or more.
I'm curious about what you are looking for with Chimera and whether there are new features you would need. We primarily focus on proteins and nucleic acids but would consider adding features to help in materials science.
Tom Goddard Chimera developer UC San Francisco
H. C. Greenwell EPSRC Fellow Centre for Computational Science Christopher Ingold Laboratories University College London 20 Gordon Street London WC1H 0AJ United Kingdom --------- Date: Thu, 21 Oct 2004 10:54:31 +0100 Subject: Chimera materials visualization wish list From: "H. Chris Greenwell" <h.greenwell@ucl.ac.uk> To: Thomas Goddard <goddard@cgl.ucsf.edu> Hi Tom, I realise that in my earlier e-mail I have only outlined general points, and therefore have only addressed part of your email. Therefore, if I could compile an initial wish-list of new features (apologies if some of this is already in the code, as I am still getting to know it) it might look something like: 1) To be able to import the common file types we use in our simulations. These are: (a) the LAMMPS 2001 output files (.pos), which are largely XYZ files with a time step and an atom type number related to, (b) the LAMMPS 2001 data file (.dat/.lmp), which contains the atom information (masses), coordinates and bonding. (c) the LAMMPS 2003 data/position files, which I think are a little different but will be used by us in the near future. (d) the Accelrys Biosym coordinate (.car) file, topology (.mdf) file and trajectory (.arc/.trj) files. Dave kindly put together a little utility for the .car/.mdf file. 2) To be able to remap any atoms that have moved outside the cell, within the cell boundaries when visualized. 3) To be able to view the cell boundaries as a dashed/solid/no line. 4) To be able to replicate the viewed model in XYZ directions so that motion across cell boundaries can be followed. 5) To be able to turn the replicated unitcells into one supercell (we have an utility to do this with .car/.mdf) . 6) To be able to display and select atoms on the basis of forcefield atom types. 7) To be able to render crystalline structures as polyhedra. 8) After visualizing, or performing molecule/atom translations/deletions/insertions etc, to be able to export as the file types listed in (1) and also the Atom Eye (.cfg) format. Many thanks Chris H. C. Greenwell EPSRC Fellow Centre for Computational Science Christopher Ingold Laboratories University College London 20 Gordon Street London WC1H 0AJ United Kingdom