Hiking in the energy landscape in sequence space: a bumpy road to good folders

With the help of a simple 20 letters, lattice model of heteropolymers, we investigate the energy landscape in the space of designed good-folder sequences. Low-energy sequences form clusters, interconnected via neutral networks, in the space of sequences. Residues which play a key role in the foldability of the chain and in the stability of the native state are highly conserved, even among the chains belonging to different clusters. If, according to the interaction matrix, some strong attractive interactions are almost degenerate (i.e. they can be realized by more than one type of aminoacid contacts) sequence clusters group into a few super-clusters. Sequences belonging to different super-clusters are dissimilar, displaying very small ($\approx 10%$) similarity, and residues in key-sites are, as a rule, not conserved. Similar behavior is observed in the analysis of real protein sequences.