Chapter 41Flexible Ligand-Protein Docking

This chapter discusses how to calculate binding energetics from structure [1], and then presents a protocol for flexible docking in solution using metadynamics [2].

Predicting Binding Energetics from Structure

The relationship between the binding affinity constant, KK , and the binding free-energy change, ΔG∘\Delta G^\circ , is given by

ΔG∘=−RTln⁡K,\Delta G^\circ = -RT \ln K, where RR is the gas constant, and TT is the temperature in Kelvin.

The binding free energy ΔG∘\Delta G^\circ can be written as

ΔG∘=ΔH∘−TΔS∘,\Delta G^\circ = \Delta H^\circ - T \Delta S^\circ, where ΔH∘\Delta H^\circ is the enthalpy change and ΔS∘\Delta S^\circ is the entropy change.

To take into account the changes in heat capacity, ΔCp\Delta C_p , that occur during protein-ligand binding reactions in water, the following expression for ΔG∘\Delta G^\circ is used instead:

ΔG∘=ΔHR∘−TΔSR∘+ΔCp(T−TR−TlnTTR),\Delta G^\circ = \Delta H_R^\circ - T \Delta S_R^\circ + \Delta C_p \left( T - T_R - T \ln \frac{T}{T_R} \right), where ΔHR∘\Delta H_R^\circ and ΔSR∘\Delta S_R^\circ are the enthalpy and entropy changes at the reference temperature TRT_R and TT is the temperature of interest.

The parameters ΔHR∘\Delta H_R^\circ , ΔSR∘\Delta S_R^\circ , and ΔCp\Delta C_p can all be calculated by modeling changes in accessible surface area (ASA), which is defined as the surface traced out by a sphere, with radius equal to that of a solvent molecule, when rolled over the surface of the molecule of interest. This can also be formulated in configuration space as the Minkowski sum of the protein surface and the sphere, namely, P⊕Br={p+b∣p∈P,b∈Br}P \oplus B_r = \{p + b \mid p \in P, b \in B_r\} , where PP is the protein and BrB_r is a ball of radius rr . This is a classical problem in computational geometry (see [3]).

Both ΔCp\Delta C_p and ΔHR∘\Delta H_R^\circ can be computed from the changes in ASA:

ΔCp=1.88ΔAap−1.09ΔApol,\Delta C_p = 1.88\Delta A_{ap} - 1.09\Delta A_{pol} , and

ΔH∘=−35.3ΔAap+131ΔApol,\Delta H^\circ = -35.3\Delta A_{ap} + 131\Delta A_{pol} , where ΔAap\Delta A_{ap} and ΔApol\Delta A_{pol} denote the changes in apolar and polar ASA upon binding.

The parameter ΔS∘\Delta S^\circ consists of the following three terms:

ΔS∘=ΔSsolv∘+ΔSconf∘+ΔSmix∘,\Delta S^\circ = \Delta S_{solv}^\circ + \Delta S_{conf}^\circ + \Delta S_{mix}^\circ , where ΔSsolv∘\Delta S_{solv}^\circ denotes the contributions from restructuring of solvent, and ΔSconf∘\Delta S_{conf}^\circ denotes the contributions from changes in conformational degrees of freedom of the backbone and side-chain groups, and ΔSmix∘\Delta S_{mix}^\circ denotes the contributions from changes in translational, rotational, and vibrational degrees of freedom upon binding. All ΔSsolv∘\Delta S_{solv}^\circ , ΔSconf∘\Delta S_{conf}^\circ , and ΔSmix∘\Delta S_{mix}^\circ can be indirectly calculated from ASA. See section 2 in Murphy [1] for more details.

Flexible Docking in Solution Using Metadynamics

Overview of Metadynamics

A metadynamics is “a dynamics in the space of collective coordinates that are evolved with a standard restrained molecular dynamics (MD) supplemented by a history-dependent potential” [2]. More precisely, each metadynamics run is an MD run in which a set of harmonic restraints on collective variables Sα(r)S_\alpha(r) is imposed, and evolved using the following modified Lagrangian form:

H+∑α[12Mα(dsαdt)2+12kα(sα−Sα(r))2]+VG(s,t),H + \sum_\alpha \left[ \frac{1}{2} M_\alpha \left( \frac{ds_\alpha}{dt} \right)^2 + \frac{1}{2} k_\alpha (s_\alpha - S_\alpha(r))^2 \right] + V_G(s, t) , where sαs_\alpha is an auxiliary variable representing the value of the restraints, MαM_\alpha is the fictitious mass, and kαk_\alpha is the coupling constant. The potential VG(s,t)V_G(s, t) is constructed as a sum of Gaussians centered on the values of sα(t)s_\alpha(t) :

VG(s,t)=w∑t′=T,2T,…t′<texp⁡{−[sα−sα(t′)]22δs2},V_G(s, t) = w \sum_{t'=T, 2T, \dots t' < t} \exp \left\{ -\frac{[s_\alpha - s_\alpha(t')]^2}{2\delta_s^2} \right\} , where ww and δs\delta_s are the height and the width of the Gaussians and TT is the time interval in between two Gaussians added.

The metadynamics method has the following two features:

  1. The reduction of search dimensionality by choosing the collective variables SαS_\alpha of the metadynamics.
  2. The possibility of the free energy surface (FES) calculation from the summed Gaussians.

Application of Metadynamics in Flexible Docking

The following metavariables are used in the metadynamics docking:

  1. The angle between the line connecting the centroids of receptor and ligand, and the principal axis of inertia of the ligand.
  2. The ligand-active site distance or the ligand-centroid of the receptor distance.

These two metavariables are illustrated in figure 41.1. Other variables, such as the number of ligand-enzyme hydrogen bonds, hydrophobic contacts, and coordinating water molecules, were

Figure 41.1

Two metavariables in the metadynamics docking. Credit: John MacMaster and Bruce Donald.

Parameters for the distance metacoordinate:

Gaussian height w=0.48 kcal/molew = 0.48 \text{ kcal/mole}
Gaussian width s=0.4 Angstroms = 0.4 \text{ Angstrom}
metadynamics step length T=3 psT = 3 \text{ ps}
coupling constant k=600 kJ/(mol Angstrom2)k = 600 \text{ kJ/(mol Angstrom}^2)
(ficticious) mass M=150 umaM = 150 \text{ uma}

Parameters for the angle metacoordinate:

Gaussian height w=0.48 kcal/molew = 0.48 \text{ kcal/mole}
Gaussian width s=0.1 rads = 0.1 \text{ rad}
metadynamics step length T=3 psT = 3 \text{ ps}
coupling constant k=1600 kJ/(mol rad2)k = 1600 \text{ kJ/(mol rad}^2)
(ficticious) mass M=400 kJ fs2/(mol rad2)M = 400 \text{ kJ fs}^2/(\text{mol rad}^2)

Figure 41.2

Table 41.1The overview of results

Goal Method of Assessment
(usually) Compare Prediction
to Observation
Simulated
Process (time)
Protein-Ligand System Used Undocking
(3 ns)
Docking
(3 ns)
Undocking
(6 ns)
Undocking
(7.2 ns)
Unspecified
(8 ns)
A B C
#1 Metacoordinates of deepest FES well vs. X-ray structure ligand “close to” “slightly shifted” X “around” “around”
RMSD of heavy atoms within 6 A of ligand NONE reported 0.3 Angstroms X 0.4 Angstroms 0.4 Angstrom
-6 vs. -6.5 -6 vs. -6.5 -5 vs. -5.8 -8 vs. -7.1 -20 vs. < -11
#3 Prediction vs. the literature value X X +1 vs +0.7 X X
#4 Compare to FES from 2D umbrella sampling “satisfactory” 3ns vs. 15ns X X X

A: β\beta -trypsin/benzamidine; B: β\beta -trypsin/chlorobenzamidine; C: immunoglobulin McPC-603/phosphocholine (PC); D: cyclin-dependent kinase 2 (CDK2)/staurosporine). Credit: John MacMaster and Bruce Donald.

discarded because either they are claimed by [2] to lack general applicability or they were not able to distinguish between different basins.

The following four protocols were used to test the metadynamics docking method:

  1. Finding the Docking Geometry Compare the coordinates of the ligand in the global minimum of the FES with the crystallographic structure. Calculate the average root-mean-square deviation RMSD of the heavy atoms in the ligand and protein between the experimental geometry and the configuration from the metadynamics simulations.
  2. The ΔGbinding\Delta G_{\text{binding}} Calculation The ΔGbinding\Delta G_{\text{binding}} is computed from the difference between the free energy of the minimum and the deeper free energy of the unbound state.
  3. The ΔΔGbinding\Delta\Delta G_{\text{binding}} Calculation.
  4. The Calculation of the Whole Docking FES The whole docking FES is calculated by a bidimensional umbrella sampling [4] using the same angle and distance variables.

Results

In Gervasio et al. [2], the following four docking cases are used to test the metadynamics docking method: β\beta -trypsin/benzamidine, β\beta -trypsin/chlorobenzamidine, immunoglobulin McPC-603/phosphocholine (PC), and cyclin-dependent kinase 2 (CDK2)/staurosporine. The choices of the metadynamics parameters are shown in figure 41.2. The overview of results is shown in table 41.1.

References

  1. [1] K. P. Murphy. Predicting binding energetics from structure: looking beyond DeltaG. Med Res Rev., 1999 Jul; 19(4):333–9.
  2. [2] F. L. Gervasio, A. Laio, and M. Parrinello. Flexible docking in solution using metadynamics. Journal of the American Chemical Society 127;8 (2005 Mar 2):2600–2607.
  3. [3] T. Lozano-Perez. Spatial planning: A configuration space approach. IEEE Transactions on Computers C-32;2 (February 1983):108–120.
  4. [4] Philip W., Shankar Kumar, and Maximiliano Vasquez Payne. Method for free-energy calculations using iterative techniques. Journal of Computational Chemistry 17;10 (1998):1269–1275.