You are here

Lammps or RosettaSurface

4 posts / 0 new
Last post
Lammps or RosettaSurface


I hope I placed this question in the proper place. If not please let me know.

I am attempting to computationally dock peptides/proteins to metal oxides etc..What is the most approbriate code to use RossettaSurface or Lammps? At the moment, I have installed PyRosetta and Lammps. Since there is a learning curve using these codes, I would like to select the best one to meet my needs. With LAMMPS, I believe there is versatility (i.e. alter atom types etc..); but I don't know about RosettaSurface..

Any advice in this area would be welcomed...



Post Situation: 
Fri, 2016-06-10 10:13

First off, a caveat: I'm not really familiar with LAMMPS, so I can't give a detailed comparison.

One consideration is that the bulk of Rosetta development is toward soluble proteins in a fully aqueous environment. There is comparatively little development of protocols which involve protein/surface interactions. Particularly, the energy functions which predict how well proteins interact with the surface have comparatively little development. Therefore, if you do a comparison between RosettaSurface and a program that has a number of dedicated people involved in optimizing the program toward protein-surface interactions, Rosetta is probably not going to be as accurate.

However, the difference in approaches between Rosetta (Monte Carlo) and LAMMPS (Molecular Dynamics) means that there are cases where Rosetta may be the better choice, even if the prediction accuracy is not as refined. This goes back to the fact that accurate molecular dynamics tends to take long runs. If you're attempting to look at a large number of conditions the speed at which Rosetta can test conditions might be a benefit, even if it's not a precise in any given condition. Additionally, Rosetta has relatively efficient ways of handling and evaluating mutations, so if you want to do design in the context of surface mutation, Rosetta may be a better choice, as opposed to a molecular dynamics program, where evaluating mutations might involve laborious setup and lengthy runs.

That said, the two are not mutually exclusive. People have had success (e.g. with enzyme design) using Rosetta to design a protein or narrow things down to a small set of possible starting conditions, and then using molecular dynamics to further refine/evaluate those designs or conditions. That is, you use the speed of Rosetta to quickly evaluate a large number of possiblities, and then use a "more accurate"(*) molecular dynamics protocol to do a final evaluation.

(* In the particular cases I'm thinking of, "more accurate" was better handling of backbone dynamics and flexibility, rather than energy function evaluation.)



Fri, 2016-06-17 10:49

Hi, I am a new user trying to learn the RosettaSurface.  Can anyone help me out to understand how to setup a model for such as the "" in the "demos/public/" ? There are three lines or three points in the file of "" : first point is the surface center-- that is a simple for me to follow; the second line and the third line in the are very confusing me to follow! How to get both point/vector information? by any softwares? or even by manually measurement? 

Tue, 2018-09-25 18:53


This is Mike Pacella, I developed much of the code specific to RosettaSurface and use it regularly in my research. Rocco is definitely correct in pointing out that the energy function used in surface docking is not as well developed as the energy function used for soluble proteins. However, in my opinion, you would be facing a similar problem with LAMMPS. Neither LAMMPS nor Rosetta (nor any package that I'm aware of) use an energy function that has been optimized for protein-surface interactions. LAMMPS will be more accurate at modeling the surface component of the system and Rosetta will be more accurate at modelling the protein component. As far as modeling the interaction between a protein and a surface goes, there really hasn't been a good benchmark put forth to test various methods. In fact, developing such a benchmark is something I am dedicating the bulk of my time in lab to. So if you have some experimental data that you think could make for a good benchmark system I'd definitely be interested in helping and if you just want help running the code I'd be happy to help as well! Feel free to contact me at 





Mon, 2016-06-20 11:55