Computational Drug Discovery and Design

explicit solvent during the simulation phase of the project, a supe-

rior approach for introducing ions is to first use a program like

CIonize, available in VMD, to place ions around the protein–drug

system of interest prior to immersion in a water box. CIonize places

ions according to local electrostatic potential of the system, elim-

inating the additional equilibration time needed to allow randomly

placed ions to localize within these energetically favorable regions

(seeNote 12). Following placement of ions proximal to the pro-

tein–drug system and immersion in explicit water, bulk ions can be

added at the desired salt concentration, taking into account the

numbers of ions already placed and ensuring the final system charge

is neutral (seeNote 13).

3.3 Parameterizing
the Protein–Drug
System for MD
Simulation

Once the initial structural model has been prepared and immersed

in a realistic solvent environment, parameterization of the system

must be addressed. Biomolecular force fields for describing the

protein component of the system are readily available, given that

the protein does not contain nonstandard residues. Commonly

employed protein force fields include those from the AMBER

(http://ambermd.org) and CHARMM (http://mackerell.

umaryland.edu/charmm_ff.shtml) families. For the example sys-

tem in this chapter, CHARMM36 [28, 29] is used (seeNote 14).

Obtaining parameters to describe the drug component of the

system is less straightforward. A typical approach is to employ a

generalized force field, such as the Generalized AMBER Force

Field (GAFF) [30] or the CHARMM General Force Field

(CGenFF) [31], which attempts to provide parameter coverage

for the numerous chemical substructures commonly encountered

in small molecules. Unfortunately, due to the complex structural

nature of many drug compounds, generalized force fields do not

always contain parameters that reasonably describe all parts of the

molecule. Any missing parameters must be derived ab initio,

according to the standard protocol of the generalized force field

they will be combined with (seeNote 15).

For the example system in this chapter, CGenFF is employed to

describe the drug molecule. An online tool is available through

ParamChem (https://cgen.paramchem.org) that performs auto-

mated atom typing and assignment of parameters and charges

available in CGenFF based on analogy of the molecule’s structure

to the ensemble of substructures covered by the force field

[32, 33]. The tool works best when supplied with a MOL2 file,

which contains information on connectivity and bond order; the

Schro ̈dinger Maestro program [34] is recommended for generating

a correctly formatted MOL2.

Importantly, the identification of parameters for the molecule

by ParamChem does not necessarily guarantee the suitability of

those parameters to accurately reproduce the molecule’s dynamical

behavior. A penalty score is provided along with each identified

Molecular Dynamics Simulations of Protein-Drug Complexes 251