Abstract Details


SubCav - Tool for Subpocket Comparison and Alignment

Tuomo Kalliokoski1, Tjelvar S.G. Olsson2, Anna Vulpetti3
1Lead Discovery Center GmbH, Dortmund, Germany
2Cambridge Crystallographic Data Centre, Cambridge, UK
3Novartis Institutes for Biomedical Research, Basel, Switzerland
Fragment-Based Drug Discovery (FBDD) is a widely applied approach in both industry and academia. In FBDD small and weakly active fragments are used as the starting points for the larger and highly active lead compounds. FBDD relies mostly on structural information gained from X-ray crystallography experiments. As the number of protein structures both in public domain and proprietary databases is constantly increasing, there is a strong incentive to utilize this source of data more efficiently.

The chemical space can be better sampled with fragments than with large compound libraries. However, it is not feasible to consider all commercially available fragments experimentally as it has been estimated that there are about 400,000 different fragments currently commercially available. Therefore, there has been considerable interest to generate computational tools for the rational fragment library design which enable creating of smaller more focused sets of fragments. One way to analyze the fragment binding environments in proteins is to use binding site similarity methods.

Two binding sites might be different when comparing them in whole, but they might still bind the same fragments if they share suitable subpockets. Well-known examples are the similar sulfonamide- and trifluoromethyl-binding subpockets between cyclooxygenase-2 and the isoenzymes of the carbonic anhydrase family which bind the same functional groups in comparable orientations. Information about shared subpockets can be therefore used in fragment-based drug design to suggest new fragments or to replace existing fragments within an already known compound.

A novel computational method called SubCav is described which allows the similarity searching and alignment of subpockets from a Protein Databank (PDB)-wide database against a user-defined query. Subpockets are defined to be the protein environments within 4.5 Å distance from any of the fragment atoms. Ligands are fragmented using the BRICS-scheme. SubCav is based on 3D pharmacophoric fingerprints combined with a subpocket alignment algorithm.
Validation was performed using a set of non-redundant PDB complexes sharing the same fragments (3,394,572 pairs). Subpockets were first aligned by their fragments to get the pairs that could be aligned successfully. Two subpockets were defined to be aligned if the Root-Median-Square Deviation (RMSD) of the fragments was less or equal than 1.5 Å and the ratio of the overlapping pharmacophoric feature was from 0.5 to 1.0. SubCav was able to produce to correct alignment in 73-85 % of these cases depending the on the overlap criteria. The method also successfully aligned pairs which were missed by the fragment-based alignment due to the small differences in fragment conformations.

Additionally, a prospective study was done using the adenine fragment of phosphomethylphosphonic acid adenylate ester (ACP) bound to Heat Shock Protein 90 (HSP90) to illustrate the use of the method for FBDD. SubCav retrieved relevant subpockets from the PDB including those with different folds. Bioisosteric replacement of an adenine to a pyrazole could be suggested from the alignment between adenine subpocket of ACP bound to HSP90 and pyrazole subpocket of a carbamate compound bound to Escherichia coli DNA gyrase B. This replacement was confirmed by a HSP90 structure that had a pyrazole containing ligand bound to it.

The method can also be used to analyze subpockets inside a protein family to facilitate drug design and to rationalize compound selectivity which was illustrated with a study on histone methyl-transferases’ binding sites which all bind the S-adenosylmethionine (SAM) co-factor, which fragments into three parts. Hierarchical clustering based on the SubCav similarities between subpockets was rationalized from the visual inspection of the binding sites. The analysis could be used to design novel mimetics for different fragments of SAM.

Return to Programme