The RS126 set of proteins was used to check that the alignments generated by our method could reproduce previous results for PHD, and DSC. PHD (run in cross-validation mode) increased in average Q3 accuracy by 1.9% from 71.6 to 73.5%, over the published accuracy. The accuracy obtained here for DSC improved by 1.0%. However, the published value of 70.1% for DSC was for a full jack-knife test, whereas our test utilised all the data. These results confirm that the automatic method used to build multiple alignments in this study is appropriate for secondary structure prediction.