$beta_2$ -Adrenergic receptor ($beta_2$ AR) is a G protein-coupled receptor that is highlyexible and able to recognize a wide range of ligands through its conformational variations. Active and inactive conformations revealed by recent crystallographic experiments do not provide a complete dynamic picture of the receptor, especially in the binding site. In this study, molecular dynamics (MD) simulation through a residue-based coarse-grained model is used as an alternative and e cient method to explore a wider conformational search space. The system was composed of $beta_2$AR embedded into a 1-palmitoyl-2-oleoyl-phosphatidylcholine membrane bilayer with surrounding water. A total of 6 s of simulation at constant NPT was performed for a system of 6868 coarse-grained beads. The system reached equilibrium at around 0.1s. The overall 3-dimensional structure was well preserved throughout the simulation. Local residue-baseductuations were in good agreement with fully atomistic MD simulations. Four distinct snapshots were selected and reverse-mapped to all-atom representations with around 65,000 atoms. Each reverse-mapped system was later subjected to 100 ns of MD simulation for equilibration. Root mean square deviation clustering analysis yielded distinct receptor conformers for the binding site regions, which were suggested to be alternative representations of the binding pocket and thus were proposed as plausible targets in docking-based virtual screening experiments for the discovery of novel antagonists.