SSM vs. others: 1COL: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
 
1COL:A (197 residues)
COLICIN *A (C-TERMINAL DOMAIN) (PORE-FORMING DOMAIN)
11 helices were used for SSE matching.
 

1.  V A S T    (server)

The picture below (Figure 1COL:A-1) shows comparison of Ca-alignment length obtained from SSM and VAST. As seen from the picture, the set of close neighbours include just 3 chains (except 1COL:A itself), those are 1COL:B, 1A87 and 1CII. SSM and VAST agree remarkably well for all PDB entries presented in the Figure, except may be for the very last ones (1A1W and 1A1Z, SSM gives somewhat longer alignments).


  Figure 1COL: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.
 

The same quality of agreement between SSM and VAST is also seen in the obtained RMSDs (Figure 1COL:A-2). In this particular example, SSM and VAST criteria for balancing the alignment length and RMSD do obviously coincide.


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

As a direct consequence of good agreement between SSM and VAST seen in Figures 1COL:A-1 and 1COL:A-2, the match indexes of SSM and VAST alignments are very close to each other (cf. Figure 1COL:A-3), even closer than alignment lengths and RMSDs themselves.


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

P-values of SSM and VAST alignments, shown in Figure 1COL:A-4, demonstrate high degree of similarity as well, however SSM gives a considerably lower P-values for the closest structural neighbours.


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

Z-scores of SSM and VAST Ca-alignments (Figure 1COL:A-5) show a reasonably good quantitative agreement except for the two last entries.


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

 

 

2.  C E (Combinatorial Extension)    (server)

Comparing to VAST, Combinatorial Extension gives much more structural neighbours for 1COL:A (cf. Figure 1COL:A-6). From CE results, one can distinguish highly similar (PDB entries 1-3, same as those for VAST), similar (entries 4-750) and remote neighbours. All of them are reproduced by SSM, however SSM offers shorter Ca-alignments for remote structural neighbours. A somewhat oddish thing is that CE reports 204 residues aligned for the very first match 1COL:A-1COL:B, although there is only 197 residues in each 1COL chain.


  Figure 1COL: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.
 

Comparison of Figures 1COL:A-7 and 1COL:A-6 reveals that in the region of similar structural neighbours (4-750) SSM gives smaller RMSDs at about-same alignment lengths as those of CE. For remote structural neighbours, SSM favours smaller RMSDs to longer Ca-alignments, as compared to CE. As appears, this gives less spiky results; the largest RMSD spike in Figure 1COL:A-7 produced by CE is for PDB entry 1LXL.


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

The match index of SSM and CE alignments shows a very good agreement (cf. Figure 1COL:A-8). It may be seen from the Figure that SSM makes slightly better alignments for close structural neghbours, in accordance with the results of visual inspection of Figures 1COL:A-7 and 1COL:A-6. An interesting point is that the RMSD spike, dominating the Figure 1COL:A-7, is not reflected in the match index, being compensated by a relatively small spike in the alignment length (also seen in Figure 1COL:A-6).


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

Z-scores of SSM and CE alignments, shown in Figure 1COL:A-9 show generally good agreement. SSM Z-scores are a bit more scattered than those of CE. It is interesting to note that for remote structural neighbours, Z-scores from CE look cut at Z=3.7 . This should be partially attributed to a limited number of significant digits in CE web-output, however it could also be a property of CE algorithm. In this particular example of 1COL:A, Z-scores of CE and SSM for 1COL:A are very close quantitatively, while we found that in most cases Z-scores from CE are twice lower than those from SSM (here the factor of 2 in Figure 1COL:A-9).


  Figure 1COL:A-9.
Z-scores corresponding to matches shown in Figure 1COL: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 1COL:A-10). It aligns all but two residues for 1COL:A itself and the closest structural neighbour 1COL:B. As seen from Figure 1COL:A-10, SSM and DALI generally agree in the length of Ca-alignments, however DALI tends to longer alignments for remote structural neighbours.


  Figure 1COL: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 Figure 1COL:A-11 suggests, longer DALI's alignments come in exchange for higher RMSDs, as compared to SSM. As seen from the Figure, RMSDs from DALI are almost always higher than those from SSM, some of them are probably higher then reasonable (5Å and more).


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

Figure 1COL:A-12 makes it clear that in spite of differences between SSM and DALI alignments, visible in Figures 1COL:A-10 and 1COL:A-11, the principal quality of alignments, produced by both servers, is very much the same. Indeed, match index of SSM alignments is nearly coinciding with that obtained from DALI.


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

DALI's Z-scores (Figure 1COL:A-13) are higher than those from SSM for highly similar structures, and lower for the remote ones. Z-scores from both DALI and SSM show very similar trends, although differ in figures. We also compare Z-scores from DALI with minus logarithm of SSM's P-values (shown by black line in Figure 1COL:A-13), because in most cases they agree better. As seen from the Figure, a good agreement in the overall trend is indeed there, however Z-scores from DALI are higher than -lg(P) from SSM.


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

 

 

4.  D E J A V U    (server)

DEJAVU gives sonsiderably shorter Ca-alignments than SSM (cf. Figure 1COL:A-14). Most of DEJAVU's output give very remote structures, which is in quite a difference from VAST, CE, DALI and SSM.


  Figure 1COL: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.
 

It is therefore expectable that DEJAVU produces shorter RMSDs, which is fully confirmed by data in Figure 1COL:A-15. The Figure makes an impression that DEJAVU imposes high penalties for RMSDs higher than 2Å. RMSD's from SSM are considerably higher however stay in a reasonable range (less than 5Å).


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

The match index shows a much better agreement than alignment lengths and RMSDs (cf. Figure 1COL:A-16). It is therefore clear that the principal quality of SSM and DEJAVU alignments is equally good, however servers differ substantially in balancing the compromise between alignment length and RMSD.


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

P-values, given by SSM and DEJAVU alignments, are shown in Figure 1COL:A-17. As seen from the Figure, SSM and DEJAVU generally agree in the assessment of statistical significance of the results. Overall, DEJAVU gives a considerably higher statistical significance of matches than SSM.


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

Z-scores of SSM and DEJAVU, presented in Figure 1COL:A-18, show a slightly better agreement than P-values (cf. Figure 1COL:A-17). Z-scores from DEJAVU are almost twice higher than those from SSM.


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

 

 

5.  Conclusion

For all presented results, there is a remarkable agreement in the principal quality of 3D Ca-alignments, as measured by match index. However, the balance between alignment length and RMSD is different for different servers. In this respect, SSM best agrees with VAST. DALI and CE results are also in a reasonable agreement with SSM, while DEJAVU shows the most discreapancy.

VAST and SSM demonstrate a very good agreement in the lengths of Ca-alignment, RMSDs, P-values and Z-scores. CE tends to longer alignments at higher RMSDs, however reasonably agrees with SSM in Z-scoring. DALI and SSM show better agreement for similar structures. With growing dissimilarity, DALI seem to tend to longer Ca-alignments at higher RMSDs, as compared to SSM. Z-scores from DALI and SSM have similar trends but differ in absolute values. DEJAVU gives shorter alignments and lower RMSDs and finds mostly remote structural neighbours.