SuMo: A Tool for Protein Function Inference Based on 3D Structures Comparisons - Identification of Ligand Binding Site and Protein-Protein Interaction Area - page 10

Table 1.1 (continued)

New object data set Original object data set

pdb code Sumo score Objects pdb code Sumo score Objects

2HXG 9.28 20 iso 2JFO 3.49 7 iso

1AW2 9.28 19 iso 2H4L 3.49 7 iso

1SUX 9.27 19 iso 1YA7 3.48 7 TEP

1JX1 9.26 22 iso 1YAU 3.48 7 TEP

1R2R 9.25 19 iso 1D6M 3.45 6 iso

2VEN 9.24 19 iso 2VA6 3.45 6 AEP

16/61 Protease : 26,2% 14/61 Protease : 22,9%

13/16 serine protease: 81,3% 8/16 Serine protease: 50%

recovered on the first 50 hits which gives a percentage of 26.2%, but 13 on 16 serine
protease were recovered with a percentage of 81.3%. The new set gives better results
that the previous one with a percentage of recovery about 22.9% of protease and
only 8 on 16 serine proteases recovered on the 50 first hits. Difference between the
two SuMo object sets is clearly establish on this run because the first 13 lines
contain serine proteases for the new set and only 4 entries at the top of the list for
the old set. In fact, deeper analyses of the result from the new set of SuMo objects
give a better answer. Only 3 serine proteases were not recovered on the first 50 hits:
1SBC (79 hits), 1CEA and 1BF9 are not present. The 1CEA (Mathews et al. 1996 )
structure is the non covalent complex of the recombinant kringle 1 domain of human
plasminogen. This structure has 7 domains, one PAN, 5 kringle and one peptidase
domain, the resolved structure have only the end on the PAN and the kringle 1
domain. Even the structure is displayed on the PDB with serine protease EC code;
the protease domain is not present in the PDB file. For the second entry “no present”
with the PDB code 1BF9, the same conclusion can be made. The structure is the N
terminal EGF like domain from human factor VII and can be divided in three
domains, Gla EGF1, EGF2 and Serine protease. The PDB structure resolved by
NMR gives the EGF1 domain whereas the other domains are absent. Therefore, the
new SuMo objects allow to recovered 13 of the 14 true serine proteases thus giving
a recovery percentage of 92.9%.
The analysis of the common object or signature is displayed for all serine protease
on Table 1.2 . The signature was compiled by residue and the number of common
objects with the query structure (1AFQ) is given in the table. A value number of 1
means that all objects of a residue were recovered. The serine protease made the
peptidase reaction with three catalytic residues: H57, D103 and S195. The H57 and
the D102 were found in the 14 “true” serine proteases, only the S195 was not recov-
ered in the 1EZX structure. One of the advantage of the method is the capability to
analyse common residues in discontinuous sequence and not only in 2D like classi-
cal approach. The 1SBC model has quite nothing in common but the three catalytic
residues, it is possible to make a pattern of recognition only based on this three resi-
dues to determine if it is a serine protease. In this case, SuMo software can be used