SuMo: A Tool for Protein Function Inference Based on 3D Structures Comparisons - Identification of Ligand Binding Site and Protein-Protein Interaction Area - page 14

to detect secondary site or “moon lighting” sites of proteins. The global analysis the
14 serine proteases shows conserved area and empty band showing that in there area
there are no common objects with 1AFQ (Kashima et al. 1998 ) .

1.3.2 Test of New Set on Lectins

Lectin proteins have a specific binding capability for oligosaccharide and particularly
glycoprotein. To make a comparison with the previous SuMo work, the same lectin
set was chosen. We choose a set of 90 3D models of lectin extracted from the Protein
Data Bank (PDB). In this set 18 lectins are not able to bind oligosaccharide (designed
hereafter as false lectin). The 2PEL (Banerjee et al. 1996 ) structure is the peanut
lectin whose structure was solved by X-ray crystallography at 2.25 Å resolution.
The LAT ligand is a lactose molecule and the structure has been co-crystallized with
two lactose molecules bound in two different sites LAT and LBT. The reference set
of objet was defined as all objects around 6.5 Å of the LAT ligand site. This distance
selects 16 objects on 2PEL model, which are distributed on the 12 residues: D80,
A82, D83, G103, G104, Y125, N127, E129, S211, L212, G213 and G214.
The comparison was performed on a double cpu Xeon type dual core and the run
took less than 1 min of CPU time. The 6 best first hits were peanut lectins with different
ligands under reference: 1QF3, 1CR7, 2PEL, 2TEP, 1BZW, 1CIW. The number of
matching objects ranges from 16 to 12, the L212 and the carbon alpha of the S211
were not recovered for 1BZW and 1CIW structures. The next hits give concanavilin
structure (PDB code: 1NLS) with 9 objects. Most hits have 5 or 6 objects in common
with the query and belong to the concanavilin family. The program classified 93 hits
with 16 to 3 objects. The analysis reveals that only 2UU8 with 7 common objects is
present and the rest of the 17 false lectins are not present in the listing. This result
compared favorably with the same query on the same server from the previous set
of object reveal better hits. Indeed, by using the previous set 14 of the 18 false lectins
were present in hits, 7 has more than 5 common objects and 7 have less 4 common,
only 4 models are not in the list (Table 1.3 ). The false lectins not recovered are:
1IOA, 1DQ2, 1ENQ and 1QFD. It is interesting to notice than 2UU8 have 5 objects
on the previous set and 7 with the new one. 2UU8 have common object dispatched
on 6 residues. Here is the list of matching pairs of residue on 2PEL vs. 2UU8: A82/
A207, D83/D208, G103/G227, G104/R228 (carbon alpha), Y125/Y12 and N127/
N14. It is interesting to notice than the carbon alpha of the G104 of the 2PEL lectin
is matching with the R228 of the false lectin 2UU8. The weight of “carbon alpha”
object is minimal in order to not favor proteins with the same fold, so the score
2UU8 is about 4.44 versus 8.05 for 2PEL and are hits number 20 / 93. Among the
12 residues, 6 has been paired explaining the “good” hit, but the mutation on gly-
cine 104 in arginine on the binding has probably disabled the capability of this
protein to bind with an oligosaccharide.