Note: this parameterization has been superseded by beta_nov15.
An overview of this energy function optimization effort is available here.
For most protocols (those that use getScoreFunction to set the protocol score function), the flag -beta_july15 will load this version of the beta energy function.
For RosettaScripts protocols, the flag -beta_july15 must be provided, and the following scorefunction declaration must be made:
The most significant change in beta_july15 is the addition of an anisotropic polar solvation model for polar sidechain atoms. This uses Phil Bradley's lk_ball model, where -- for polar sidechain atoms -- virtual water sites are placed in ideal geometry, and the solvation model uses a weighted sum of the distance to the virtual water site and the distance to the heavyatom when calculating how desolvated a polar sidechain atom is. The net effect is increased desolvation cost when placing an occluding atom near one of these virtual waters.
LK parameters (fa_sol)
Following introduction of the lk_ball anisotropic polar solvation model, the LK parameters were refit following our energy function optimization scheme. Initially, several manual modifications were manually made:
Then, using the LK-ball anisotropic solvation model, the DGFREEs of all atom types were refit following our optimization criteria. As with the original LK paper, agreement to liquid-vapor transfer free energies was used as a strong constraint on optimization.
LJ parameters (fa_atr/fa_rep)
The LJ parameters (both LJ_RADIUS and LJ_WDEPTH) for all atom types were refit during energy function optimization. Three strong constraints were employs to keep these parameters physically realistic:
Electrostatics dielectric model (fa_elec)
A sigmoidal dielectric model was used in place of the default distance-dependent dielectic. Three parameters control the short range dielectric, long range dielectric, and the fade between the two. Initially these parameters were chosen to match the original dielectric as closely as possible; they were subsequently included as free parameters in optimization.
Hydrogen bond strengths
Per-donor and per-acceptor hydrogen bond weights were optimized following our optimization criteria. A weak tethering constraint in our "energy function goodness" target function was used to keep these weights uniform.
Several minor modifications were made to the energy function as well: