MM-PB(GB)SA and Umbrella Sampling (US) are both methods fro free energy calculations. However, for a ligand binding problem with (the surface of) a protein, which one is better, when compared to experiments?
Some literature reference will be highly appreciated.
For comparison of US to expt., I found this paper: Article Effective Estimation of Ligand-Binding Affinity Using Biased...
However, I was unable to find anything similar for MM-PBSA vs expt.
Since MM/PB(GB)SA relies on an implicit solvent and doesn't cover entropy, US is certainly more accurate. However, as always, accurate and performance barely come together, so the real question is, which method is fast enough for your needs and also sufficiently accurate.
You might also want to have a look at FEP for accurate absolute binding energies.
Article Absolute Binding Free Energy Calculations: On the Accuracy o...
Norman Geist: Thank you for the reply. However, I was wondering if it is really true that US is costlier than MM-PB(GB)SA methods, in terms of CPU hours and/or wall-clock hours. FEP, I agree, is extremely costly.
Martin Klvana: Thank you for the paper. It is quite interesting.
I'd say so. In MMPB(GB)SA you only need the complex simulstion. In US you need many (although rather short) MD simulations at different umbrellas, that is, different restrained distances between ligand and binding site and the PMF (WHAM) may not converge as quickly.
MM/PBSA provides a computationally efficient method for large-scale calculations. It is quite good and widely used in drug screening. In many cases, their performance could be as good as more accurate methods. More detailed simulation-based methods may have many unexpected problems and you may not have very good experience with it, although they are theoretically rigorous in many aspects.
FEP/TI would be more accurate (within the limitations of the chosen force field) for getting the absolute binding free energy. The problem with umbrella sampling is that you need to construct a continuously connected pathway between the bound and unbound (in water) states, which would be important if you are interested in the pathway and kinetics of the ligand binding/unbinding process. However, if you're interested only in the binding free energy at a specific site, perturbation based methods would be preferred, and definitely more accurate than MM/PB(GB)SA type methods, which do not take into account the explicit solvation and entropic factors (except for certain approximations).
- Badges
- Science topic
- Similar topics
- Chemistry
- Theoretical Chemistry
- Computational Chemistry
- Molecular Dynamics