home / work / projects / Ligand—protein interactions

Hydrophobic complementarity in ATP—protein complexes

Pyrkov et al. (2007). — PROTEINS 66(2), 388–398.

The problem:

Since structural characterization of protein—ligand complexes is experimentally demanding, special attention is now paid to in silico techniques, among which molecular docking is the most commonly used. However, it has been observed that while docking algorithms are usually able to yield correct conformation as compared to the experimental structures, they often fail to rank it top among other, wrong conformations.

A common way to overcome this problem is to apply more effective scoring criteria to the final set of several docking poses to select the correct one. Usually such criteria take into account some characteristic features of the protein target, or, more rarely, of the ligands under study. The rationale for this is that such ligand- or target-specific scoring criteria implicitly account for certain intermolecular interactions, typical for that particular types of compounds, which offer certain advantage over a general-purpose scoring functions.

Objective:

Adenosine triphosphate (ATP) is a substrate of numerous biological reactions in the living cells. Therefore, detailed investigation of structural aspects of interactions between ATP and its protein targets is of a great importance for understanding the functional mechanisms of enzymes and rational drug-design. Our aim was:

• to investigate the intermolecular interactions which are specific to ATP—protein complexes;
• to propose an efficient scoring criterion for ranking the solutions of ATP docking.

Fig. 1. Characteristic intermolecular interactions in ATP—protein complexes revealed by the analysis of 50 high-resolution (< 3 Å) structures, available in Protein Data Bank.

Results:

Analysis of 50 ATP—protein complexes of high-resolution (< 3 Å), available in Protein Data Bank reveals that the most common are hydrophobic and stacking interactions of the adenine with protein and hydrogen bonds engaging its amino group (Table I).

Based on these findings a scoring criterion „ATP-SCORE“ was developed as a linear combination of the adenine—protein interactions (with the 40 complexes of the training set). Hydrophobic interactions provided the major contribution to the scoring criterion. The value of this term was determined as the product of two factors (see Figure 2):

• embedding (S2+S3)/(S1+S2+S3);
• matching (S2’’+S3’’)/(S2’+ S2’’+S3’+S3’’).

Fig. 2. Hydrophobic interactions between the adenine moiety of ATP and the binding site. A. Hydrophobic (brown) and hydrophilic (blue) properties of the ATP-binding domain of sarcoplasmic reticulum Ca-ATPase and the ATP molecule. The arrow points to the adenine binding pocket. B. A scheme, illustrating the embedding factor in the hydrophobic complemetarity. C. A scheme, illustrating the matching factor in the hydrophobic complemetarity.

To evaluate the stacking and hydrogen bonding interactions we applied simple geometric criteria, see [Pyrkov et al., 2007] for more details. The proposed scoring function also showed quite good efficiency on the test set of the rest 10 complexes (Table II).

Fig. 3. Distribution of the correct and misleading docking solutions of the 40 complexes of the training set over the ATP-SCORE and „goldscore“ values. A substantial separation of these two types of docking poses is observed with scoring by ATP-SCORE as compared to the „goldscore“.

© 2003–2007
batch2k.