8

I would like to test the influence of solvation parameters in implicit solvation models and wonder which codes are freely available as standalone programs for protein folding of small proteins, and which use energy minimization approaches instead of dynamics, so I am not looking for molecular dynamics codes.

Open the way
  • 711
  • 1
  • 9
  • 15

4 Answers4

8

The most popular molecular dynamics codes are namd and gromacs, and maybe also desmond. These packages are freely available and "open source" in the sense that the source code is made available for free. Wikipedia hosts a list of software for molecular modeling which may contain other good links.

These codes, however, are also quite complex and therefore if you want to make changes of your own, e.g. to test new solvation models, you might spend quite a bit of time just trying to understand them. If ease of use trumps speed (probably not completely if you're looking at time-scales for folding), you might want to look at mmtk, which lets your program custom molecular dynamics simulations in Python.

If you're interested in speed, but not all the bells and whistles of a full-fledged simulation package, you may also be interested in molecular dynamics libraries such as OpenMM or mdcore (disclaimer: mdcore is my own software project). These will, however, require quite a bit more programming on your own side.

Pedro
  • 9,573
  • 1
  • 36
  • 45
  • Thanks a lot! I forgot to tell a detail, I have updated the question. By the way, what is the difference between openmm and your mdcore? – Open the way Feb 04 '12 at 13:28
  • @flow: Actually, you've changed your question completely. Both namd and gromacs do energy minimization. Since the expensive and complicated part is still the evaluation of the non-bonded forces, you can also use a library such as OpenMM or mmtk to compute these and do the minimization yourself. The difference between OpenMM and mdcore is that we solve the same problem in very different ways :) – Pedro Feb 04 '12 at 13:44
  • Excellent! I did not know about energy minimization with gromacs. Now i will try to find if it has been used for protein folding in that way. Also openmm is interesting since i think it has been implemented on gpus. Which problem are you referring to? Lo puedes contar? ;) – Open the way Feb 04 '12 at 14:55
  • This is a bit off-topic, but the "problem" is molecular dynamics on shared-memory architectures such as multi-core CPUs, the Cell/BE architecture or GPUs. While OpenMM and mdcore differ significantly in how they parallelize computations, they are more similar to each other than to namd or gromacs, which both follow a distributed-memory MPI-based parallelization paradigm. – Pedro Feb 04 '12 at 15:07
  • Ok, I think I get your idea. My main concern is that if you spend too much time in some new memory management technique or similar, things could change very fast in the forthcoming years and your project could become obsolete – Open the way Feb 06 '12 at 11:40
  • by the way, the link to "mdcore" does not work – Open the way Feb 06 '12 at 11:40
  • @flow: Sure, things can always change, but I think that shared-memory parallelism, e.g. multi-cores and GPUs, are here to stay at least for a while, as they are the currently the only solution driving Moore's Law. – Pedro Feb 06 '12 at 11:59
5

Besides Pedro's suggestions, you could also look at Quantum Espresso, which among other things allows for Car-Parrinello computations. It is (or at least, it was) written in Fortran 90.

Community
  • 1
Francesco
  • 151
  • 3
3

The term "energy minimization" usually refers to the minimization of the potential energy. For protein folding simulations, you really need to minimize free energy. If you run plain energy minimization on an unfolded protein, you will pretty soon get stuck in a local minimum. So you probably do want molecular dynamics, and use some protocol such as simulated annealing in order to get to progressively lower energies. See the answer by pedro for software suggestions.

Community
  • 1
khinsen
  • 1,731
  • 11
  • 8
  • yes, I meant free energy, but I want to avoid molecular dynamics – Open the way Feb 06 '12 at 18:14
  • 1
    If you want a free energy, which involves entropy and thus a sampling of conformational space, you have two options: use a simple potential (typically a harmonic one) and do an analytical computation, or use a non-trivial and realistic potential and do sampling. For sampling, lots of techniques have been developed, but they all boil down to molecular dynamics or Monte-Carlo, with molecular dynamics much better established for the specific case of proteins. – khinsen Feb 15 '12 at 10:40
  • yes, I agree with your general answer completely, but my problem now is to find the right, ready-to-use related codes – Open the way Feb 15 '12 at 14:55
1

There is also an interesting program called foldit which looks like a puzzle game with a nice GUI but in the background it actually studies protein folding.

erhanturan
  • 426
  • 3
  • 6