Although accuracies reported by different methods are not always directly comparable because of different data sets being used, there is a clear trend of a slow but steady improvement over the past 24 years.
← all excerpts
Sixty-five years of the long march in protein secondary structure prediction: the final stretch?
2
—
—
The sentences
The fraction of constructed three-dimensional structures with a correct fold (RMSD < 6 Å [ 133 ]) for all 182 725 40-residue fragments in a data set of 1199 proteins is small but highly significant (16.3% by predicted φ and ψ angles and 19.1% by predicted θ and τ angles).