Sequence identity levels

Can one push the limits of the method so as to use models with less and less sequence identity with the target, i.e. models obtained by threading methods, with sequence identity levels between 15 and 25% The answer seems to be yes (Jones, 2001), with newer methods capable of solving problems with sequence identity around 20% (Claude et al., 2004 G. Labesse, personal commimication Abstract in the GTBio Meeting, June 2004, Lyon, France Keegan and Wynn, 2007). 

If significant similarities between uncharacterized and annotated sequences are found, the transfer of annotation is straightforward at high identity levels (> 40%). Below that sequence-identity level, the establishment of firm sequence-function relationships can start to present some problems. The signal starts to become blurred at 20 to 30% identity, the so-called Twilight Zone [12,13]. If identity levels are even lower, the region is called the Midnight Zone (fig. 3.2). For identity levels over 30%,... 

Suprafamilies a group of homologous enzymes that catalyze different overall reactions which do not share common mechanistic attributes and do not share a high level of sequence identity (< 20%) but are metabolically linked such as in successive reactions. Active-site residues may be conserved but perform different functions in family members.Tables 16.2 and 16.3 list known superfamilies and suprafamilies with some of their family members. 

---