Our recent paper was featured on F1000:
Recent results in the field of protein folding have showed that atomistic molecular dynamics (MD) simulations can correctly fold small proteins, providing insights on their structures as well as on their folding pathways. Those simulations, however, remain computationally expensive. MD simulations are deterministic: once the initial positions and velocities have been set, the trajectory for the protein under study is derived by numerically solving Newton's second law with tiny time steps, hence the long simulation times. If additional information (such as "forming a hydrophobic core") could be added to steer the simulations, it is expected that they could be sped up significantly. However, proper inclusion of such additional information requires caution as it needs to satisfy the thermodynamic principles of Boltzmann's law. The authors have developed a proper statistical mechanics framework for this purpose, they have applied this framework to successfully fold twenty small proteins, including ubiquitin, a millisecond folder. Inclusion of this framework into folding simulations using MD is shown to yield up to five orders of magnitude faster computational folding times than traditional MD.