Free Energy Perturbations in Ribonuclease T1Substrate Binding. A Study of the Influence of Simulation Length, Internal Degrees of Freedom and Structure in Free Energy Perturbations

Abstract

We have studied the reliability of free energy perturbation calculations with respect to simulation protocol and simulation length in a real biological system, the binding of two different ligands to wildtype Ribonuclease T ₁ (RNT1) and to a mutant of RNT1 with Glu-46 replaced by Gln (RNT1-Gln46). The binding of the natural substrate 3′ GMP has been compared with the binding of a fluorescent probe, 2-aminopurine 3′ mono phosphate (2AP3′MP). These simulations predict that the mutant binds 2AP3′MP better than 3′GMP. Four complete free energy perturbations were performed that form a closed loop of four free energy differences, which should sum up to zero. This could be used as a tool for searching for systematic errors that are not detected by standard forward ↔ backward perturbations. The perturbation between 2AP3′MP and 3′GMP is quite straightforward and similar to what has been done by other groups. The perturbation between Glu46 and Gln46 is much more complex, involving as many as twelve atoms and a change of charge. This perturbation needs much longer simulation time, 500-600 ps, than used in free energy perturbations before. The increased simulation time is needed both to reach an equilibrium and to include several phases of fluctuations of the observed parameters in the production run. The extremely long simulation time is not such a severe problem as much of the work might be done on several different machines in parallel and cheap workstations are excellent for these calculations. Problems may also occur with values of the coupling parameter Λ close to 0 or 1, due to the high mobility of atoms as well as insertion/deletion in a previously unoccupied space involved in the perturbation.