We present a retrospective analysis of CASP3 threading predictions, applyin
g evaluation and assessment criteria used at CASP3. Our purpose is twofold,
First, we wish to ask whether measures of model accuracy are comparable be
tween CASP3 and CASP2, even though they have been calculated differently. m
e find that these quantities are effectively the same, and that either may
be used to compare model accuracy. Secondly, we wish to assess progress in
fold recognition by comparing the numbers of CASP2 and CASP3 models that cr
oss specific accuracy thresholds. We find that the number of accurate model
s at CASP3 drops sharply as the targets become more difficult, with less ex
tensive similarity to known structures, exactly the pattern seen at CASP2.
CASP3 teams do not seem to have predicted accurate models for targets of gr
eater difficulty, and for a given difficulty range the best CASP3 models se
em no more accurate than the best models at CASP2, At CASP3, however, we fi
nd greater numbers of accurate models for medium-difficulty targets, with e
xtensive similarity to a known structure but no shared sequence motifs. Thr
eading methods would appear to have become more reliable for modeling based
on remote evolutionary relationships. Published 1999 Wiley-Liss, Inc.dagge
r.