Druggability of a target protein is an important question in drug discovery. A reliable and physically relevant measure of druggability can help determining risks associated with pursuing a given target protein. Furthermore, a comprehensive analysis of druggability of a target can help identifying novel sites and alternate inhibition mechanisms.
To this extent, experimental NMR and X-ray and computational screening methods are utilized for druggable binding site identification. Protein pockets that bind a wide range of fragments or organic solvent molecules in these screening experiments usually coincide with known druggable sites.
Based on these ideas, we developed a unbiased simulation based approach to assess druggability of target proteins with known structures. Our approach involves simulations of the target in presence of a set of small organic molecules (probes) with diverse physiochemical properties. Probes are selected to be small (four non-hydrogen atoms) so that they can diffuse fast and explore small and even transient pockets in simulations. This property also helps sampling large number of binding events and enables reaching equilibrium in relatively short simulation times.
Simulation trajectories are analyzed to calculate probe enrichment on protein surface and pockets using a grid based approach. Enrichment grids are converted to probe binding affinities using inverse Boltzmann relation. Binding sites are identified by locating clusters of high affinity probe binding spots. Druggability index for a binding site is calculated by considering the affinities of seven or eight probe molecules (28 to 32 non-hydrogen atoms), which is equivalent to a drug-like molecule in size.
Details of the approach can be found in [AB12]_. We showed that probes mimic interactions of drug-like molecules, as well as substrates and inhibitors that are not necessarily drug like. Thus, the approach is suitable for assessing druggability or ligandability of a protein.