Comparative Analysis of Techniques Oriented on the Recognition of Ligand Binding Area in Proteins - Identification of Ligand Binding Site and Protein-Protein Interaction Area - page 75

Table 4.1 (con tinued)

NAD+ FMN

Low High Low High
MCC F-m MCC F-m MCC F-m MCC F-m
Cutoff 80 % 1KYQ 1GIQ 2GWL 1KQN 2NZL 1YKG 2Z98 3CB0
1TOX 1KYQ 1HDR 1HDR 3CB0 1SBZ 1OBO 1N9L
3JYO 1P1H 1N2S 1N2S 1I0R 1RLJ 2VZJ 3B9O
3M6I 1TOX 1FDV 1OBB 3B9O 1OBO 2D5M 2NZL
1TOX 2GR9 2I2F 2Q1U 2Z6I 1NOX 1MVL 1I0R
3JSA 3JSA 1KQN 1FDV 1VP8 1MVL 1B1C 1QZU
3M6I 3JYO 1OBB 2I2F 2BYC 1F4P 2XOD 2R6V
1KYQ 3KET 2Q1U 1KQN 2BYC 1EJE 1RLJ 2Z6I
3KET 3M6I 1FDV 3OX4 2Z6D 1B1C 1YKG 1VP8
SuMo 3CIN 3OX4 2GR9 2GR9 2Z6I 3CB0 1EJE 1EJE
1DHS 1TOX 1BMD 1BMD 3RH7 2 KB2 1F4P 1F4P
1GR0 1V9L 1OG3 1OG3 2ZRU 2 KB2 1VP8 1VP8
1SC6 2PH5 2JHF 2WN7 2 KB2 1NOX 1MVL 1MVL
1TOX 1SC6 1 AD3 1 AD3 3P7N 1VYR 1SBZ 1SBZ
1V9L 1DHS 1KYQ 2JHF 3A3B 3P7N 1DNL 2WQF
1ZBQ 3P2O 2WN7 1GIQ 3CB0 3BW2 1I0R 1VYR
2PH5 1GR0 3JYO 2PLA 2PR5 3A3B 2WQF 1DNL
3KET 1ZBQ 1GIQ 1EVJ 3BW2 2Z6D 1VYR 2Z98
3P2O 3KET 1EVJ 1JQ5 2Z6D 2PR5 2Z98 2NZL
PDB codes printed in boldface indicate unique matches (i.e. matches found only by a single software
package). Underlined codes indicate proteins with the highest frequency of matches for a given
classi fi cation

Ligand

For FMN, the most correctly identified binding site appeared to be in 2BYC
protein (CASTp, QsiteFinder, ConSurf).
The biological properties of the analyzed proteins are very diverse and do not
seem to correspond to binding pocket prediction accuracy. The study group includes
enzymes and transport proteins, monomers as well as complexes consisting of indi-
vidual subunits. It would be difficult to attribute identification accuracy to any common
property putatively shared by all of the presented proteins.
The 2Z6I protein was very accurately analyzed by the FOD model but posed
significant problems for QSiteFinder, CASTp, SuMo and PocketFinder. FOD
also yielded accurate results for 1N9L, contrary to SuMo and PocketFinder. This
suggests that the geometry of the binding pocket in these proteins is rather generic
and that the interaction between the ligand and the hydrophobic core plays a deci-
sive role.
An entirely different situation occurs in 2BYC which was accurately processed
by all models except FOD. It seems that in 2BYC the presence of a ligand does not
distort the protein's own hydrophobic core to an appreciable degree.
Determining the factors responsible for binding pocket prediction accuracy is a
complicated problem. Proteins which could be easily identified by geometry-based