Today we know of over 1000 protein structures, which can be classified
into approximately 120 distinct folding patterns. The database of kno
wn structures provides numerous examples of proteins that adopt very s
imilar folds, with some in each folding class having similar sequences
. But there are also examples of proteins with similar structures that
share no obvious sequence similarity. Thus among the 60 000 known ami
no acid sequences, there must be many that adopt the 120 known folds b
ut cannot be identified based on sequence relationships alone. It is t
he goal of inverted protein structure prediction to determine whether
an amino acid sequence adopts a known structure. Here, we review the r
ecent, rapid progress in inverted structure prediction. The power of t
his new generation of methods is that, instead of looking for similari
ty in sequences, they attempt to match one-dimensional sequences direc
tly to three-dimensional folds.