Figure 1

Comparison of performance of AlignHUSH method to HHSearch and PRC. A) The sensitivity and error rate values for both AlignHUSH and HHSearch are plotted in this figure. The sensitivity of AlignHUSH is better than HHSearch or PRC at almost all error rates. The 'no_sec', 'no_hyd' and 'no_neigh' are variants of AlignHUSH procedure without use of secondary structure, hydrophobic and neighboring column information respectively. B) Alignment accuracy of the three methods that have been examined in detail in the main text. The alignment accuracy given in this plot corresponds to the 'developer score' defined in the main text. The three methods are comparable as far as the accuracy using developer score is concerned. C) The alignment accuracy of the three methods using the 'modeller score' defined in the main text. The performance of AlignHUSH is slightly better than that of HHSearch and PRC. HHSearch generated alignments tend to be very short and hence HHSearch has a low value for 'modeller score' alignment accuracy. D) The length of the query HMM covered by the alignment is plotted for the alignment between homologous families (two SCOP families belonging to the same SCOP superfamily). The coverage of query HMM is greater in case of AlignHUSH than HHSearch which indicates that AlignHUSH generated alignments are more informative for function annotation, since they cover almost the entire homologous region. The alignment length coverage is very similar between the PRC generated alignments and AlignHUSH generated alignments.