Hi everyone,
I have heard about different fancy methods ( e.g. thermodynamic integration, Free energy perturbation, umbrella sampling) but what if I just want to use simple MD to calculate the free energy difference, how can I do that ( I agree that will require a lot of sampling please ignore it ) ?
If you want to compute the free-energy difference between two states from simulations at these states only, you can do two things:
1. If the phase space between the simulations overlaps sufficiently, you can apply the fancy methods (e.g. free energy perturbation) in a single step. As you already guessed, this will require a lot of sampling time, because you are typically not interested in free-energy differences between states that are very similar. Or you can specifically design a single system such that it shows overlap with the states that you are interested in.
2. If you do not care about using an exact statistical mechanical method, but you want to roughly estimate the free energy difference between two simulations, then you can use end-state methods, which usually try to approximate the free energy by making a number of assumptions. For ligand binding, typical examples are MMPB/SA or some flavour of linear response (LIE, LRA).
If you want to compute the free-energy difference between two states from simulations at these states only, you can do two things:
1. If the phase space between the simulations overlaps sufficiently, you can apply the fancy methods (e.g. free energy perturbation) in a single step. As you already guessed, this will require a lot of sampling time, because you are typically not interested in free-energy differences between states that are very similar. Or you can specifically design a single system such that it shows overlap with the states that you are interested in.
2. If you do not care about using an exact statistical mechanical method, but you want to roughly estimate the free energy difference between two simulations, then you can use end-state methods, which usually try to approximate the free energy by making a number of assumptions. For ligand binding, typical examples are MMPB/SA or some flavour of linear response (LIE, LRA).
@James “Wes” Barnett Thanks for very informative answer. I have asked question just out of curiosity. I usually use thermodynamic integration or umbrella sampling for free energy of solvation calculations.
@Chris Oostenbrink Thanks for the very informative answer. you second point is really interesting. I really like to dig out some literature on that.
- Badges
- Science topic