I have tried two
members of Src tyrosin kinase family Src (2src) and Hck (2hck)
These proteins are 450 respective 438 aa long and according to CATH
classification they contain four domains in their chains:
Number in parentheses in column Another true positives means number of
retrieved true positives
|
Server |
Trivial solution found (itself) |
Non-trivial solutions found |
Rank of first false positive |
Time |
|
CE |
yes |
yes |
Last 5 |
8 |
|
Dali |
yes |
yes |
33,42 |
5 |
|
Dejavu |
no |
Yes (51) |
no |
190 |
|
Lock |
no |
yes |
6 |
16 |
|
Matras |
yes |
Yes(30) |
No |
33 |
|
Pride |
no |
Yes(1) |
2. |
Sec |
|
SSM |
yes |
Yes(6) |
no |
5 |
|
Top |
yes |
Yes(10) |
no |
11 |
|
Tops |
no |
no |
1 |
5 |
|
Topscan |
no |
no |
1 |
Sec |
|
Vast |
yes |
Yes(52) |
no |
31 |
2hck 3,0 , 438 res., 50 % SSE
|
Server |
Trivial solution found (itself) |
Non-trivial solutions found |
Rank of first false positive |
Time |
|
CE |
yes |
yes |
Far away |
ND |
|
Dali |
yes |
yes |
31 |
52 |
|
Dejavu |
no |
Yes (45) |
no |
210 |
|
Lock |
no |
yes |
6 |
16 |
|
Matras |
yes |
yes |
No |
30 |
|
Pride |
no |
no |
1 |
sec |
|
SSM |
yes |
Yes(6) |
no |
5 |
|
Top |
yes |
yes |
No |
10 |
|
Tops |
No |
no |
1 |
6 min |
|
Topscan |
no |
Yes(1 in first 20) |
2 |
Sec |
|
Vast |
yes |
yes |
No |
ND |
CE processing time is unknown, because we
used pre-computed results.
Comments to results:
CE more results for 2hck than 2src (78 to 67), but also more false positives.
The results are properly ranked. First are the structures with four matched
domains, than whole bunch of structures with two matched domains and than
structures with one matched domain ordered according to size of domains. All
CATH defined domains for these structures are covered. False positives are at
the really end.
Dali the same case as CE, perfectly ranked results.
Dejavu failed to find proteins
that contein either only SH2 or SH3 domain, but found the structures that
contain at least two of the domains in common with SRC kinases.
Lock - in neither of cases it found best results (that mean structures with
four matched domains). My explanation is, that these structures are not
contained in their database. But the program was able basically find all types
of domains, which it was supposed to find, although they were mixed with false
positives almost from beginning.
Matras similar to CE and Dali, no false positives, but also missing results
(maybe because of higher cut-off z-score)
Pride found only one hit with kinase domain, no SH3 and SH2 domains.
SSM is able to find only structures with all four domains present, not
able to find structures containing only some of the domains existing in Src
kinases.
Top no false positives, but I couldnt find between the hits any SH2 and SH3
domains. It found only kinase domain, it is not so good.
Tops can not handle with multidomain proteins, no hits
Topscan one hit to kinase domain in one case, but otherwise nothing. It can
not handle with multidomain proteins.
Vast no false positives, the same case of Dali, Matras and CE. Perfect.