Allosteric Sites Prediction

Predicting allosteric binding sites on the surface of a target protein

Q-MOL molecular modeling software provides a unique capability – a reliable identification of allosteric regulatory site on the surface of a target protein.

Here, we refer the term “allosteric site” to an existing active site, cryptic site, any surface area of a target protein.  In order for allosteric site to efficiently interact with a small molecule ligand, the allosteric site must store certain excess of structural energy.  This excess structural energy usually comes from inner structural conflicts (torsional, non-bonded, etc).  These structural conflicts are resolved upon conformational change of a protein and/or binding of specific protein binding partner.  This is the normal way for the nature to take care for thermodynamic energy balance to support specific protein-protein and protein-ligand interactions.

Case Study – human kRas protein

The crystal structure of human kRas protein (PDB 4OBE, chain A) was scanned using Q-MOL QPortal service.  The figure below represents initial analysis of surface scan results.  The data from SpecificityIndex.dat file were plotted as bar plot.  The data points (individual amino acids) forming exponential portion of the plot were selected for display in Pymol. 

The Pymol session file created by QPortal job, contains selected protein chain, and molecular objects corresponding to individual amino acids.  The amino acids, chosen from specificity plot, were displayed while the rest were hidden (see inset).

The following figure represents allosteric sites analysis:

As seen from the figure, two well-defined allosteric sites were detected on the surface of the kRas protein: site 1 and 2.  The excess surface energy was converted into probability values and displayed as spheres: small/blue – low probability, red/big – high probability.  The location of GDP binding pocket is indicated.  On the third slide of the figure, the close up of GDP binding pocket is shown together with GDP molecule.

Why this pocket is not detected?  This pocket’s protein structure environment is fully optimized.  There are no non-bonded, torsional and other unresolved molecular structure defects and interactions.  As a result, the excess surface energy is insignificant.

The GDP binding pocket is fine tuned to specifically and tightly bind GDP ligand (IC50 is low nanomolar).  Any derivative of GDP will not fit this pocket so perfectly.  And this is the main reason why all attempts to drug this pocket with non-covalent inhibitors have failed.

Design of Binding Peptide Toward the Predicted Allosteric Site

The results of allosteric sites detection could be used to design a short peptide sequence which could be a representation of putative protein binding partner.  The peptide could be also synthesized and used as probe to test interactions at the predicted binding site.

Note that generated peptide is highly optimized by force field parameters toward the corresponding binding site, and optimized peptide could be significantly different from a natural binding partner.

Here we will oversimplify the use case, and will discard such sequence design considerations as N-C termini direction, sequence permutations, grouping residues by spatial proximity, etc.

Consider predicted allosteric site 2 of kRas:

The amino acids on the Pymol objects view panel are already arranged according to the Specificity Index.  To choose sequence direction N -> C, we will simply follow Specificity Index: Val -> Glu.  The reverse order is equally possible, and should be tried under different circumstances. 

The spatial arrangement of amino acids does not unambiguously define the order of amino acids.  The constructed peptide represents one of many possible residue permutations of its sequence.

Even in this oversimplified case, the BLAST protein sequence alignment search of the constructed peptide reveals kRas protein binders as top hits:

The highlighted entries are actual protein binding partners of kRas.