In many cases, the homology that is inferred from
structure has allowed interesting functional assignments
to be made. For example, a hypothetical Saccharomyces
cerevisiae protein was found to be a triosephosphate isomerase
(TIM) barrel, the active site of which looks like
alanine racemase, and preliminary studies indicate that it
does have that biochemical activity16.However, homology
has not proved to be definitive; indeed, of the ten
structures solved by Christendat and co-workers, in no
case did structurally inferred homology alone provide a
robust functional prediction31. In several cases, common
ancestry inferred from structure has not reflected
common function; for example, Methanobacterium
thermoautotrophicum MTH538 closely resembles the
Escherichia coli response regulator CheY, but could not be
shown to have any related aspartate-kinase activity81.
Furthermore, two close homologues of unknown molecular
function — YjgF from E. coli and YabJ from Bacillus
subtilis — were both found to be similar in structure to
chorismate mutase.However, the completely different
active sites precluded the possibility of these proteins
sharing chorismate-mutase function with their structurally
similar homologue50,51. So, although structural
analysis failed to show the role of YjgF and YabJ, it was key
in allowing the researchers to realize that their homology
did not reflect similar activity. Structure determination of
M. thermoautotrophicum MTH1175 likewise showed
structural similarity to E. coli RNaseH, but did not support
a shared function between the two82.Because active
sites can occur in different contexts and can change in
homologous proteins, several automated methods have
been developed to seek similarity in active sites to predict
function83–85 or specificity86,87; however, the application of
these methods has not yet been described for the handful
of published structural genomics proteins.