Nucleoproteins belonging to the HMG-1/-2 family possess homologous dom
ains similar to 75 amino acids in length. These domains, termed HMG-1
boxes, are highly structured, compact, and mediate the interaction bet
ween HMG-1 box-containing proteins and DNA in a variety of biological
contexts. Homology model building experiments on HMG-1 box sequences '
threaded' through the H-1-NMR structure of an HMG-1 box from rat indic
ate that the domain does not have rigid sequence requirements for its
formation. Energy calculations indicate that the structure of all HMG-
1 box domains is stabilized primarily through hydrophobic interactions
. We have found structural relationships in the absence of statistical
ly significant sequence similarity, identifying several candidate prot
eins which could possibly assume the same three-dimensional conformati
on as the rat HMG-1 box motif. The threading technique provides a meth
od by which significant structural similarities in a diverse protein f
amily can be efficiently detected, and the 'structural alignment' deri
ved by this method provides a rational basis through which phylogeneti
c relationships and the precise sites of interaction between HMG-1 box
proteins and DNA can be deduced.