SSM vs. others: 1AD5:A

 
Materials from this page cannot be reproduced without permission from the authors.
Comparisons made on November 2002, using current versions of VAST, CE, DALI, DEJAVU and SSM v1.22 from 20/11/2002.
 

CONTENTS
  1. VAST
  2. CE (Combinatorial Extension)
  3. DALI
  4. DEJAVU
  5. Conclusion
 
1AD5:A (438 residues)
SRC FAMILY KINASE HCK-AMP-PNP COMPLEX
13 longest helices and 15 strands were used for SSE matching.
 

1.  V A S T    (server)

Figure 1AD5:A-1 shows the Ca-alignment lengths obtained from SSM and VAST for different structural neighbours (chosen by VAST). As seen from the picture, both SSM and VAST results allow for clear identification of highly similar (PDB entries 1-9), less similar (10-175) and two groups of remote (176-278,279-366) structural neighbours. As seen from the Figure, with the exception of a few cases, SSM results agree with those from VAST very well. Not very typically for the comparison with VAST, SSM offers offers a somewhat shorter Ca-alignments for less similar neighbours in the range of PDB entries 70-175.


  Figure 1AD5:A-1.
Length of Ca-alignment as a function of PDB entry, obtained by SSM (black line) and VAST (red line). Details of the calculations are given here.
 

As seen from Figure 1AD5:A-2, in that range Ca-alignments from VAST have higher RMSDs. Otherwise, RMSDs from SSM and VAST are in a reasonably good agreement, showing a reasonable balance with the alignment length. There are, however, a few SSM spikes in the region of PDB entries 244-249, which look unbalanced with the alignment length respecting to VAST results (meaning that VAST seems to give better RMSDs at approximately the same alignment lengths). It should be noted here that although VAST recognizes 1AD5:A correctly, it fails to produce a proper alignment with zero RMSD for it. As may be seen from the Figure, VAST aligns 1AD5:A (PDB entry 1) with RMSD equal to 0.3 Å.


  Figure 1AD5:A-2.
RMSD of Ca-alignment corresponding to data in Figure 1AD5:A-1. Details of the calculations are given here.
 

This imperfect alignment of 1AD5:A is also reflected in the VAST match index (cf. Figure 1AD5:A-3): the match index for 1AD5:A is 0.82 against 1 given by SSM. Otherwise, there is a good agreement between SSM and VAST match indexes. An average degree of the agreement (within 5%) is also observed in the region of PDB entries 244-249, where RMSDs from SSM look particularly spiky (cf. Figure 1AD5:A-2). This implies that SSM alignments for those entries are just as good as those from VAST, however they are balanced differently in respect to the alignment length and RMSD.


  Figure 1AD5:A-3.
Match Index corresponding to data shown in Figure 1AD5:A-1. Details of the calculations are given here.
 

P-values of SSM and VAST alignments, shown in Figure 1AD5:A-4, demonstrate similar trends, however SSM produces lower P-values as a rule. As may be seen from the Figure, only the borderline between highly similar and less similar structural neighbours is clearly seen in the P-values (compare with Figure 1AD5:A-1). Remote structural neighbours have only slightly lower statistical significance, as compared to the overall scattering of the results.


  Figure 1AD5:A-4.
P-values corresponding to matches shown in Figure 1AD5:A-1. Details of the calculations are given here.
 

Z-scores of VAST Ca-alignments (Figure 1AD5:A-5) are higher than those from SSM for highly similar and less similar structures. SSM and VAST agree very well in Z-scores for remote structures. both SSM and VAST Z-scoring allow for identification of highly similar structural neighbours, however SSM Z-scores, in difference of those from VAST, do not show a clear borderline between less similar and remote structural neighbours (compare Figures 1AD5:A-1 and 1AD5:A-5).


  Figure 1AD5:A-5.
Z-scores corresponding to matches shown in Figure 1AD5:A-1. Details of the calculations are given here.
 

 

 

2.  C E (Combinatorial Extension)    (server)

As may be seen from Figure 1AD5:A-6, lengths of Ca-alignments, produced by SSM, generaly agree with those obtained from CE. Both SSM and CE allow for the identification of highly similar (1-8), less similar (9-140), and two groups of remote (141-280,281-433) structural neighbours. A systematic difference between SSM and CE is seen in the region of less similar structural neighbours (PDB entries 9-140), where CE gives longer alignments. In the region of remote structural neighbours, SSM alignment lengths make a bottom line for those obtained from CE.


  Figure 1AD5:A-6.
Length of Ca-alignment as a function of PDB entry, obtained by SSM (black line) and CE (red line). Details of the calculations are given here.
 

Analysis of the Ca-alignment RMSDs (Figure 1AD5:A-7) shows that longer CE alignments come at the expense of higher RMSDs. This is see partiucularly well in the region of less similar structural neighbours (PDB entries 9-140), although the general correlation "longer alignment - higher RMSD" may be seen for almost all entries (compare Figures 1AD5:A-6 and 1AD5:A-7).


  Figure 1AD5:A-7.
RMSD of Ca-alignment corresponding to data in Figure 1AD5:A-6. Details of the calculations are given here.
 

The match index of SSM alignments seems to be slightly better (higher) than those of CE alignments for similar structures (PDB entries 9-250), although the difference is small and irregular. Overall, the match index prooves that the principal quality of SSM and CE alignments is very similar. Analysis of Figures 1AD5:A-6, 1AD5:A-7 and 1AD5:A-8 suggests that SSM and CE generally agree in the compromise between alignment length and RMSD.


  Figure 1AD5:A-8.
Match Index corresponding to data shown in Figure 1AD5:A-6. Details of the calculations are given here.
 

Z-scores of SSM are on average twice higher than those obtained from CE (cf. Figure 1AD5:A-9). The highly similar structures are more clearly identified in SSM Z-scoring. In the region of remote structural neighbours (PDB entries 141-433) SSM's Z-scores are more scattered.


  Figure 1AD5:A-9.
Z-scores corresponding to matches shown in Figure 1AD5:A-6. Details of the calculations are given here.
 

 

 

3.  D A L I    (server)

Comparing to VAST and CE, DALI produces less matches (cf. Figure 1AD5:A-10). DALI results allow for the identification of highly similar (PDB entries 1-9), less similar (10-34) and remote (35-95) structural neighbours. As seen from Figure 1AD5:A-10, SSM and DALI generally agree in the length Ca-alignments, however DALI tends to longer alignments for less similar structural neighbours. For individual structural neighbours, SSM and DALI show considerable (up to 50%) difference in the alignment length.


  Figure 1AD5:A-10.
Length of Ca-alignment as a function of PDB entry, obtained by SSM (black line) and DALI (red line). Details of the calculations are given here.
 

As seen in Figure 1AD5:A-11, there is a very good correlation between Ca-alignment RMSDs obtained from SSM and DALI. The Figure makes an impression that DALI RMSDs are almost always higher than those from SSM. While that is definitely true for the region of less similar structures (PDB entries 10-34), a comparative analysis of Figures 1AD5:A-10 and 1AD5:A-11 shows that higher RMSDs always correspond to longer alignments.


  Figure 1AD5:A-11.
RMSD of Ca-alignment corresponding to data in Figure 1AD5:A-10. Details of the calculations are given here.
 

The above conclusion is fully corroborated by data presented in Figure 1AD5:A-12. As seen from the Figure, the match indexes, calculated from SSM and DALI results, show a remarkable closeness. This indicates that the principal quality of alignments, given by these servers, is essentially the same. A comparative analysis of Figures 1AD5:A-10, 1AD5:A-11 and 1AD5:A-12 yields that SSM and DALI differ in balancing the alignment length and RMSD for individual entries.


  Figure 1AD5:A-12.
Match Index corresponding to data shown in Figure 1AD5:A-10. Details of the calculations are given here.
 

DALI's Z-scores (Figure 1AD5:A-13) are higher than those from SSM, for highly similar and less similar structures, and lower for the remote ones. Z-scores from both DALI and SSM show similar trends, although differ in figures. As a typical finding here, Z-scores from DALI agree much better with minus logarithm of SSM's P-values (shown by black line in Figure 1AD5:A-13).


  Figure 1AD5:A-13.
Z-scores corresponding to matches shown in Figure 1AD5:A-10. Details of the calculations are given here.
 

 

 

4.  D E J A V U    (server)

DEJAVU failed to recognize most closest structural neighbours, including 1AD5:A itself and gave 1FMK as the closest prototype of the input. This entry is the only highly similar structural neighbour in DEJAVU results. As seen from Figure 1AD5:A-14 most of DEJAVU's output represents remote structures, which is in quite a difference from VAST, CE, DALI and SSM. Most of Ca-alignments from DEJAVU are considerably shorter than those computed by SSM.


  Figure 1AD5:A-14.
Length of Ca-alignment as a function of PDB entry, obtained by SSM (black line) and DEJAVU (red line). Details of the calculations are given here.
 

As seen from Figure 1AD5:A-15, DEJAVU produces considerably shorter RMSDs, as compared to SSM. The analysis of spikes in Figures 1AD5:A-14 and 1AD5:A-15 suggests that there is a good correlation between alignment length and RMSD produced by the servers. It is quite clear from the Figure that DEJAVU imposes high penalties for RMSDs higher than 2Å. RMSDs from SSM are considerably higher, however stay in a reasonable range (less than 5Å).


  Figure 1AD5:A-15.
RMSD of Ca-alignment corresponding to data in Figure 1AD5:A-14. Details of the calculations are given here.
 

The match indexex, calculated from SSM and DEJAVU results, appear virtually coinciding (cf. Figure 1AD5:A-16). Given the difference in alignment lengths and RMSDs in Figures 1AD5:A-14 and 1AD5:A-15, one may therefore conclude SSM and DEJAVU differ substantially in balancing the compromise between alignment length and RMSD at similar principal quality of alignments.


  Figure 1AD5:A-16.
Match Index corresponding to data shown in Figure 1AD5:A-14. Details of the calculations are given here.
 

P-values of SSM and DEJAVU alignments are shown in Figure 1AD5:A-17. As seen from the Figure, there is a similarity in the SSM and DEJAVU assessment of statistical significance of the results, including the most pronounced spikes for individual PDB entries. P-values from DEJAVU are almost always lower than those from SSM, with a few exceptions for most remote structural neighbours.


  Figure 1AD5:A-17.
P-values corresponding to matches shown in Figure 1AD5:A-14. Details of the calculations are given here.
 

Statistical significance of matches, as given by SSM and DEJAVU Z-scoring (Figure 1AD5:A-18), represents a replica of that given by P-values (cf. Figure 1AD5:A-17). Z-scores from DEJAVU are almost always higher than those from SSM.


  Figure 1AD5:A-18.
Z-scores corresponding to matches shown in Figure 1AD5:A-14. Details of the calculations are given here.
 

 

 

5.  Conclusion

The principal quality of 3D Ca-alignments, as measured by match index, is in a good agreement between results produced by SSM and those obtained from other servers. However, the servers show a difference in solving the compromise between alignment length in RMSD. In that respect, SSM agrees well with VAST, CE, DALI, while shows a considerable difference from DEJAVU.

SSM alignments are somewhat shorter than those produced by VAST, CE and DALI, for less similar structural neighbours (alignment length in the region of 200-250, approximately 50% of 1AD5:A length). Alignments for highly similar and remote structural neighbours show better agreement. Correspondingly, RMSDs from VAST, CE and DALI are higher than those from SSM for neighbours with moderate similarity, while RMSDs for highly similar and remote structures compare relatively well. DEJAVU makes considerably shorter alignments and finds just one highly similar structural neighbour. P-values and Z-scores demonstrate similar trends however differ in numbers. This allows one to use them for rating the matches, expecting that the correspondence in the rating will be preserved, on average, across different servers.