Computational Drug Discovery and Design

details, such as formation of unexpected pores in the membrane

system, separation of the lipid leaflets, lipid bilayer thickness, deu-

terium order parameters, lateral diffusion coefficients of lipids, ion

aggregation on the protein surface, and loss of protein secondary

structure, should be assessed. If any of these properties calculated

during simulation steps is inaccurate, it may lead to major simula-

tion artifacts.

2.1.4 Generation
of Conformational
Ensemble
and Representative Protein
Structures

Selecting a representative structure from the MD trajectory frames

for analysis is one of the biggest challenges in the utilization of large

structural ensembles. The choice of the selection criteria may vary

depending on the goal of the subsequent analysis. The representa-

tive structure could be selected based on the RMSD, conforma-

tional energy, or certain known conformational changes which may

be crucial for protein function. In our protocol, the trajectories of

the MD production run for each system were monitored based on

the total conformational energy, TM6 tilt angle, and RMSD rela-

tive to the initial structure. We used an integration time step of 2 fs,

and for the analysis, simulated trajectories were saved for every 2 ps

and sampled every 300 ps. Thus, each system consists of a confor-

mational ensemble of 1000 structures for a total production run of

300 ns. The minimal energy conformation structures obtained

from the conformational ensembles were selected as the represen-

tative protein structures for the apo and agonist-bound forms of

A2AAR, which were subsequently subjected to network analysis.

2.2 Elucidation
of GPCR Allostery
Using Network
Analysis

2.2.1 Sequence
Conservation Free Energy

Sequence information of a protein family provides insight into their

functional mechanism [28]. In order to compare the conservation

pattern in class A GPCR and its specific subtype family, i.e., adeno-

sine receptor (AR), we computedΔG(GPCR)/kBT*andΔG(AR)/

kBT*, each of which is evaluated using different multiple sequence

alignment (MSA). Sequences of AR family (219 sequences) and

class A GPCR family (26,655 sequences) were collected from Uni-

ProtKB and Pfam databases, respectively. After removal of redun-

dant sequences, 208 and 24,507 sequences were considered for AR

and GPCR families, respectively. For GPCR family, sequence clus-

tering was performed using a sequence identity of at least 40% to

reduce the sequence space size, resulting in 2471 sequences. For a

given MSA of a protein family, we quantified the extent of sequence

conservation by using the following statistical free energy-like func-

tion scaled by an arbitrary energy scalekBT*[21, 29]:

ΔGi=kBT∗¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

1

Ci

X 20

α¼ 1 pi

αlogp

i

α=p

α

 2

r

ð 1 Þ

whereCiis the number of amino acid types at positionialong the

sequence,αdenotes amino acid species,piαis the frequency of an

amino acidαat the positioni, andpαis the frequency of an amino

Molecular Dynamics Approach for Investigation of GPCR Allostery 461