Target Selection Methods

Our goal is to select high quality structures from each metafold, where possible, from our Consensus Domain Dictionary (CDD) for simulation. The structures of the domains within each metafold in our CDD were examined. Where there was more than one structure available within a metafold, domains comprising an entire chain in the PDB were preferred over domains from mulitmers and multi-domain proteins. Shorter domains were preferred over larger ones with a maximum cutoff of 450 residues and high resolution crystal structures (resolution cutoff of 3 Å) over NMR structures. Potential protein domains were also selected to be self-contained and likely to fold independently. We preferred domains without cofactors (with the exception of heme, Zn and Ca), were not part of a membrane and were globular with regular secondary structure elements. Finally, for each metafold, the list of suitable protein target structures was reduced by selecting for those systems with available experimental data in the forms of NMR observables, folding and unfolding studies, drug design, or protein interaction. If at least one domain within a metafold met these criteria the domain was assigned as the fold representative for that metafold and simulated. If no domains within a metafold met all of the above criteria, a domain was assigned as the fold representative and the reasons for rejection were annotated. In the 2009 CDD, 809 metafolds had at least one domain suitable for simulation while 886 metafolds were rejected.