Folding@home: Achievements from over 20 years of citizen science herald the exascale era

Expanded ensemble predictions of absolute binding free energies in the SAMPL9 host–guest challenge

“…From this information, optimized λ -values were obtained and used for production-run EE simulations on the Folding@home distributed computing platform. 34,35…”

“…Absolute binding free energies for host-guest interactions were calculated using a double-decoupling method in which the alchemical free energies of decoupling the guest in the presence and absence of the host were computed using expanded-ensemble molecular simulations performed on the Folding@home distributed computing platform. 34,35 A threepart workflow was implemented to (1) prepare systems, (2) perform expanded ensemble simulations on Folding@home and Temple University high-performance computing (HPC) clusters, and (3) analyze the results.…”

“…Other methods such as l-dynamics 32,33 also have this ability, but EE is conceptually simpler to implement (EE is essentially an MCMC wrapper outside of the molecular dynamics integration, whereas l-dynamics requires a specialized integrator). Given this simplicity and selfcontainment in a single simulation, EE approaches are ideal for large-scale virtual screening on distributed computing platforms such as Folding@home, 34 where simulation instances are necessarily asynchronous and only loosely uncoupled to other instances. Indeed, EE was recently deployed on Folding@home 35 to screen potential inhibitors of the SARS-CoV-2 main protease as part of the COVID Moonshot initiative.…”

Markov State Model of Solvent Features Reveals Water Dynamics in Protein-Peptide Binding

“…Molecular simulations of ab initio p53 TAD peptide binding to MDM2 were performed on the Folding@home distributed computing platform − using the Gromacs 4.5.4 software package, as described by Zhou et al The AMBER ff99sb-ildn-NMR force field was used with the TIP3P explicit solvent model . Starting receptor configurations were taken from the p53-bound MDM2 receptor structure (PDB:1ycr), with unfolded and folded configurations of p53 TAD (residues 17–29) initiated at a range of receptor–ligand distances (2, 4, 6, 8, 10, and 12 Å).…”

Rescuing off-equilibrium simulation data through dynamic experimental data with dynAMMo

“…With dynAMMo, we take into account, both, simulation and experimental data to obtain a single mechanistic kinetic model. initiatives, such as Folding@home [38] or GPUGrid are more widely available allow us to access processes on the micro-to millisecond timescale, in particular when large ensembles of simulations are analyzed using Markov state models (MSMs) [39][40][41][42][43]. MSMs represent the molecular kinetics fully: the relevant structural states, their thermodynamic weights and their mutual exchange rates [44][45][46][47][48][49][50].…”