Fig. 3
From: DECIPHER: harnessing local sequence context to improve protein multiple sequence alignment
a Optimized structure matrix for pairings between helix (H), β-sheet (E), or coil (C) states. b Repeated values are grayed-out since the matrix is symmetric. After incorporating this matrix into alignment, the average improvement in Q-score on pairwise alignments was greater for distant pairs. c Alignments using the structure matrix (open symbols) showed little decline in accuracy as the number of input sequences increased relative to alignments made without structural predictions (closed symbols). Across all alignment sizes, the use of secondary structure improved Q-score (p < 1e-5 for all). Similarly, the improvement in Q-score (separation between open and closed symbols) increased as more sequences were aligned