I've been trying to figure out the concept of FAH. I read some of the older papers on the subject in the results section like No 4 for instance. This article said that different simulations make different progress, some may remain in a pseudo-stable state for long time, while others are more "productive". When simulation results are sent to the server, then a simulation which advanced the most gets picked and a new generation of simulations is based on it. A generation contains hundreds of clones, with the protein in the exact same folding state but the initialization parameters slightly different.
This article was published 4 years ago, so it was probably written 5 years ago. I beleive TINKER was still used back then. Now that most of FAH has migrated to variations of GROMACS, is the same approach still used? Every project has a Run/Clone/Gen id and the Clone and Gen part kind of hints that the methodology is not entirely dissimilar
but what exactly is a Run?In the case I'm completely wrong here, are there any articles or forum posts available that discuss how FAH is parallelized with the GROMACS technology?
Thanks in advance.