— Replica exchange is a powerful sampling algorithm and can be effectively used for applications such as simulating the structure, function, folding, and dynamics of proteins and drug design. However, Grid-based implementations of the algorithm present significant challenges due to its synchronization and communication requirements. This paper presents an asynchronous formulation of the replica exchange algorithm and the design and implementation of a Grid-based asynchronous replica exchange engine (GARE). GARE is based on CometG, a decentralized computational infrastructure for Desktop Grid environments that provides a scalable communication and interaction substrate and presents a virtual semantically specialized shared straction. It enables the dynamic and asynchronous interactions required by the algorithm to be simply expressed and efficiently implemented. The design and implementation of GARE/CometG and the replica exchange simulations that it enables are presented. Experimen...