Accuracy of alignment and structure prediction as a function of the average percent identity among input sequences. The KKA dataset was used. The alignment and structure in Rfam were assumed to be correct and the difference from them were estimated with the SPS (for assessing the alignment accuracy) and MCC (for assessing the accuracy of structure prediction). The programs used for predicting secondary structure are indicated in parentheses. The percent identities were calculated from the reference alignments. The curves were fitted using a cubic spline.