Figure 1.
Local covariation identifies alignment shift errors.
(A) A synthetic alignment was created for covariation analysis. Each of the 7 positions (columns) in the alignment contained a random assortment of 3 residues. A subset of the sequences (rows) in the alignment were then shifted for positions 2, 3, 4, 5, and 6 one position to the right. Position 1 and position 7 were not shifted. (B) A matrix of all pairwise covariation scores for the unshifted synthetic alignment where darker grey represents higher covariation calculated as in Materials and Methods. All positions randomly assort compared to one another; thus, panel B represents the background covariation for the synthetic block. Jalview conservation scores are also shown for each position. (C) Matrix of covariation scores where 3% of sequences (6 of 200) are shifted for positions 2–6. Covariation increases between all shifted positions, but does not increase between unshifted positions 1 and 7 and any of the unshifted positions. Conservation scores remain unchanged. (D) Matrix of covariation scores like panel C, except 5% of sequences are shifted. Covariation scores increase further between shifted positions, but unshifted positions show scores comparable to background as in panel A. Conservation scores remain unchanged.
Figure 2.
Alignments with high local covariation found in alignment databases.
Each alignment in the Conserved Domain Database [6] and BAliBASE 3 [4] is represented by a single circle. Alignments are partitioned by the number of sequences and the number of regions of high local covariation. A region of high local covariation is defined as a local covariation peak greater than or equal to 2.0. Alignments with two adjacent regions of high local covariation are coloured blue. Regions that contain three or more contiguous regions of high local covariation are coloured red. (A) Analysis of all conserved domains (cd) in the Conserved Domain Database (CDD). (B) Analysis of all alignments in BAliBASE 3.
Figure 3.
Realigning serine protease using LoCo.
(A) Region of high local covariation and good conservation from alignment BB40047 from the BAliBASE 3 benchmarking dataset [4]. Five highlighted sequences do not show conservation of the disulphide bonded cysteine shown structurally in panel C. (B) Realignment of region from panel A using local covariation as a guide. (C) Structural validation of the alignment from panel B built in Cn3D [39]. Positions homologous to those shown in panels A and B are coloured by identity; the conserved disulphide bond is highlighted in orange. (D) Neighbour joining tree of high local covariation segment shown in panel A. Potentially misaligned sequences (indicated by arrows) cluster in a clade joined to the remainder by a long branch. (E) Neighbour joining tree based on realigned segment in B shows realigned sequences no longer cluster together as expected by the phylogenetic relationship of the organisms.
Figure 4.
Local covariation identifies structural alignment error in BAliBASE 3 alignment of tRNA Synthetases (BB30002).
Each panel shows a structure alignment built with Cn3D [39] with the corresponding local covariation histogram from LoCo below. (A) Structure alignment of the tRNA synthetase subfamilies from BAliBASE 3. Structures are coloured by fit and the maximum local covariation value (2.8) implies a misalignment exists. (B) Realignment of misaligned structure from panel A reduces local covariation (maximum peak 1.8). Both panels A and B look very similar which explains why misalignment was missed during BAliBASE manual curation process. (C) Structure alignment of only the misaligned region of Prolyl tRNA Synthetase subfamily from panel A. Structures are coloured by identity such that conserved residues are red. Local covariation maximum is 3.0. (D) Realignment of panel C to minimize local covariation. Minimizing local covariation produces marked improvement in both the structure alignment quality and sequence conservation.
Figure 5.
Local covariation identifies active site residues.
(A) Screenshot from the LoCo tool showing the region of high local covariation from BAliBASE 3 alignment BB11003. BB11003 is an alignment of four paralogous oxireductases with similar structure. The local covariation peaks highlight four positions in the sequence alignment which are coloured in blue in panel B. Two active site residues from structure 1AD3, E333 and F335, are emphasized in the sequence alignment. (B) Structure alignment of residues shown in panel A made in PyMOL [40]. The region of high local covariation is highlighted in blue; structure 1AD3 is emphasized with dark blue. The NAD cofactor from structure 1AD3 is drawn in red. Important binding residues E333 and F335 from 1AD3 are rendered in sticks representation.