bioinformatics - Loria
BIOINFORMATICS
Vol. 00 no. 00 2012 Pages 1–10
Fast Protein Structure Alignment using Gaussian Overlap Scoring of Backbone Peptide Fragment Similarity David W. Ritchie1∗, Anisah W. Ghoorah1,2 , Lazaros Mavridis3 , and Vishwesh Venkatraman4 1
` Inria Nancy, 615 Rue du Jardin Botanique, 54600 Villers-les-Nancy, France Universite´ de Lorraine, LORIA, 54506 Nancy, France 3 University of St. Andrews, St. Andrews KY16 9AJ, Scotland, UK 4 Norwegian University of Science and Technology, Høgskoleringen 5, Trondheim, Norway 2
SUPPLEMENTARY MATERIAL
∗ to
whom correspondence should be addressed
c Oxford University Press 2012.
1
Ritchie et al.
Supplementary Figure 1. (continued)
2
Protein Structure Alignment
Supplementary Figure 1. Cartoon representations of the secondary structure assigments calculated by Stride, DSSP, and Kpax for ten example structures selected from the ten pairs of structures from Fischer et al. (1996). Secondary structure elements are coloured according to α-helix: blue; 3-10-helix: mauve; β-strand: green; turn: yellow; loop/coil: red. By construction, Kpax detects only helices and strands, and it calls all other regions as loop/coil.
3
Ritchie et al.
Supplementary Figure 2. (continued)
4
Protein Structure Alignment
Supplementary Figure 2. Cartoon representations of the superpositions calculated by CE, Sheba, TM-Align, and Kpax for the ten alignment examples of Fischer et al. (1996) drawn in the same order as Table 1 (see Table 1 for RMSDs). In each image, the fixed (reference) structure is shown with all α-helices and β-strands drawn in red. The superposed structures are shown with α-helices in blue and β-strands in green. For all structures, 3-10 helices are shown in mauve, and turns are shown in yellow. For each row, the close correspondence of the positions of the blue α-helices and green β-strands shows that each alignment algorithm is correctly recognising and superposing each pair of protein folds. All images are drawn using Stride SSE assignments.
5
Ritchie et al.
Supplementary Table 1. The RMSDs between the calculated positions of the moving structure with respect to a fixed reference structure for each of the ten pairs of structures from Fischer et al. (1996) when superposed by CE, Sheba, TM-align, and Kpax.∗
Sheba
CE Kpax
TM-Align
Sheba Kpax
TM-Align CE Kpax
1fxiA/1ubqA 1tenA/3hhrB 3hlaA/2rheA 2azaA/1pazA 1cewI/1molA 1cidA/2rheA 1crlA/1edeA 2simA/1nsbB 1bgeB/2gmfA 1tieA/4fgfA
2.46 0.59 2.46 4.03 2.07 1.93 3.53 1.08 8.34 1.21
4.76 0.72 3.96 0.84 1.49 1.15 1.71 0.82 5.44 1.24
2.83 0.37 5.94 1.65 0.57 0.49 2.88 0.70 4.59 0.17
4.57 0.27 6.30 3.55 1.31 1.00 3.48 0.90 4.95 0.69
1.78 0.89 3.62 2.90 1.70 1.55 1.80 0.60 4.69 1.01
3.83 0.37 1.11 2.58 0.92 0.75 1.00 0.42 1.72 0.38
Average
2.77
2.21
2.04
2.70
2.05
1.31
˚ units) between the coordinates of the moving structure when calculated by a specific pair of structural This table lists the Cα RMSD (A alignment methods with respect to a fixed reference structure. For example, the first element of the first row shows that the Cα RMSD ˚ The final row of between the superposed positions of ubiquitin (chain A of PDB code 1ubq) when calculated by CE and Sheba is 2.46 A. the table shows the average RMSD between the coordinates of the moving structures as calculated by each pair of alignment methods. For ˚ which example, the last element of the final row shows that the average RMSD between the superpositions of TM-Align and Kpax is 1.31 A, is less than that of all other pairs of alignment methods. In other words, on average, the Kpax alignments are more similar to the TM-Align alignments than the alignments calculated by either CE or Sheba. ∗
Supplementary Figure 3. Close-up view of the structural alignments calculated by TM-Align and Kpax for the pair 1fxiA/1ubqA (ferrodoxin/ubiquitin). In each image, the α-helices and β-strands of the fixed reference structure (here, ferrodoxin) are drawn in red. SSEs are drawn using Stride SSE assignments. Black arrows highlight particularly tight regions of the Kpax alignment compared to the TM-Align alignment.
6
Protein Structure Alignment
Supplementary Figure 4. Structural alignment comparison between TM-Align and Kpax using six low sequence identity structures from Sippl and Weiderstein (2008), drawn in the same style as Supplementary Figure 3.
7
Ritchie et al.
Supplementary Figure 5. Structural alignment comparison between TM-Align and Kpax using six difficult structures from Gerstein and Levitt (1998), drawn in the same style as Supplementary Figure 3.
8
Protein Structure Alignment
Supplementary Figure 6. Close-up view of the structural alignments calculated by TM-Align and Kpax for the pair 4aahA/1gofA from Supplementary Figure 5 (drawn in the same style as Supplementary Figure 3).
9
Vol. 00 no. 00 2012 Pages 1–10
Fast Protein Structure Alignment using Gaussian Overlap Scoring of Backbone Peptide Fragment Similarity David W. Ritchie1∗, Anisah W. Ghoorah1,2 , Lazaros Mavridis3 , and Vishwesh Venkatraman4 1
` Inria Nancy, 615 Rue du Jardin Botanique, 54600 Villers-les-Nancy, France Universite´ de Lorraine, LORIA, 54506 Nancy, France 3 University of St. Andrews, St. Andrews KY16 9AJ, Scotland, UK 4 Norwegian University of Science and Technology, Høgskoleringen 5, Trondheim, Norway 2
SUPPLEMENTARY MATERIAL
∗ to
whom correspondence should be addressed
c Oxford University Press 2012.
1
Ritchie et al.
Supplementary Figure 1. (continued)
2
Protein Structure Alignment
Supplementary Figure 1. Cartoon representations of the secondary structure assigments calculated by Stride, DSSP, and Kpax for ten example structures selected from the ten pairs of structures from Fischer et al. (1996). Secondary structure elements are coloured according to α-helix: blue; 3-10-helix: mauve; β-strand: green; turn: yellow; loop/coil: red. By construction, Kpax detects only helices and strands, and it calls all other regions as loop/coil.
3
Ritchie et al.
Supplementary Figure 2. (continued)
4
Protein Structure Alignment
Supplementary Figure 2. Cartoon representations of the superpositions calculated by CE, Sheba, TM-Align, and Kpax for the ten alignment examples of Fischer et al. (1996) drawn in the same order as Table 1 (see Table 1 for RMSDs). In each image, the fixed (reference) structure is shown with all α-helices and β-strands drawn in red. The superposed structures are shown with α-helices in blue and β-strands in green. For all structures, 3-10 helices are shown in mauve, and turns are shown in yellow. For each row, the close correspondence of the positions of the blue α-helices and green β-strands shows that each alignment algorithm is correctly recognising and superposing each pair of protein folds. All images are drawn using Stride SSE assignments.
5
Ritchie et al.
Supplementary Table 1. The RMSDs between the calculated positions of the moving structure with respect to a fixed reference structure for each of the ten pairs of structures from Fischer et al. (1996) when superposed by CE, Sheba, TM-align, and Kpax.∗
Sheba
CE Kpax
TM-Align
Sheba Kpax
TM-Align CE Kpax
1fxiA/1ubqA 1tenA/3hhrB 3hlaA/2rheA 2azaA/1pazA 1cewI/1molA 1cidA/2rheA 1crlA/1edeA 2simA/1nsbB 1bgeB/2gmfA 1tieA/4fgfA
2.46 0.59 2.46 4.03 2.07 1.93 3.53 1.08 8.34 1.21
4.76 0.72 3.96 0.84 1.49 1.15 1.71 0.82 5.44 1.24
2.83 0.37 5.94 1.65 0.57 0.49 2.88 0.70 4.59 0.17
4.57 0.27 6.30 3.55 1.31 1.00 3.48 0.90 4.95 0.69
1.78 0.89 3.62 2.90 1.70 1.55 1.80 0.60 4.69 1.01
3.83 0.37 1.11 2.58 0.92 0.75 1.00 0.42 1.72 0.38
Average
2.77
2.21
2.04
2.70
2.05
1.31
˚ units) between the coordinates of the moving structure when calculated by a specific pair of structural This table lists the Cα RMSD (A alignment methods with respect to a fixed reference structure. For example, the first element of the first row shows that the Cα RMSD ˚ The final row of between the superposed positions of ubiquitin (chain A of PDB code 1ubq) when calculated by CE and Sheba is 2.46 A. the table shows the average RMSD between the coordinates of the moving structures as calculated by each pair of alignment methods. For ˚ which example, the last element of the final row shows that the average RMSD between the superpositions of TM-Align and Kpax is 1.31 A, is less than that of all other pairs of alignment methods. In other words, on average, the Kpax alignments are more similar to the TM-Align alignments than the alignments calculated by either CE or Sheba. ∗
Supplementary Figure 3. Close-up view of the structural alignments calculated by TM-Align and Kpax for the pair 1fxiA/1ubqA (ferrodoxin/ubiquitin). In each image, the α-helices and β-strands of the fixed reference structure (here, ferrodoxin) are drawn in red. SSEs are drawn using Stride SSE assignments. Black arrows highlight particularly tight regions of the Kpax alignment compared to the TM-Align alignment.
6
Protein Structure Alignment
Supplementary Figure 4. Structural alignment comparison between TM-Align and Kpax using six low sequence identity structures from Sippl and Weiderstein (2008), drawn in the same style as Supplementary Figure 3.
7
Ritchie et al.
Supplementary Figure 5. Structural alignment comparison between TM-Align and Kpax using six difficult structures from Gerstein and Levitt (1998), drawn in the same style as Supplementary Figure 3.
8
Protein Structure Alignment
Supplementary Figure 6. Close-up view of the structural alignments calculated by TM-Align and Kpax for the pair 4aahA/1gofA from Supplementary Figure 5 (drawn in the same style as Supplementary Figure 3).
9
