Kohlhoff, K.J. et al. Nat. Chem. 6, 15–21 (2014).

G protein–coupled receptors (GPCRs) control important signaling pathways in the body and are the targets of over one-third of medical drugs. Predicting how GPCRs convert between active and inactive forms can provide insights into ligand binding and suggest new modes of activation, but detailed molecular dynamics simulations require modeling thousands of atoms and are feasible only on supercomputers. Kohlhoff et al. now take advantage of excess cloud computing power at Google to carry out an atomic-scale simulation of the GPCR β2-adrenergic receptor, which is involved in obesity, diabetes and asthma. They use idle time on servers with Google Exacycle and then use Markov state modeling to compile shorter simulations into a two-millisecond-long model, revealing a number of parallel conformational pathways and breaking the previous simulation length record by an order of magnitude.