CAPRI Critical Assessment of Prediction of Interactions
CAPRI docking targets CAPRI-01 (unbound/unbound) Lactobacillus HPr kinase - B. subtilis HPr CAPRI-02 (unbound/bound) bovine rotavirus VP6 - Fab CAPRI-03 (unbound/bound) flu hemagglutinin - Fab HC63
Docking Algorithms Flow Molecular Surface representation Critical points selection Matching of critical points Clustering and Scoring Active site knowledge
Molecular Surface Representation Budda Connoly surface for grid distance function calculations Shuo points as critical points Kely Connoly surface as critical points
Critical Points Selection Budda Volume function is computed for each shuo point and knobs and holes are selected. Kely Volume function is computed for each point and local minimum and maximum of knobs and holes are selected.
Matching of Critical Points Both algorithms use Pose-Clustering technique: For each triangle of receptor compute the transformation to each ligand matching triangle. Cluster transformations. Score the results. Complexity: O(m 3 n 3 )
Reference Frame Selection Budda Each reference frame is defined by one critical point (knob, hole) and two it’s neighbors. The signature includes triangle sides length, the angles a 1, a 2 and torsion angle w for AB and AC. Kely Each reference frame is defined by 2 points (knobs,holes) and their normals. The signature information includes the distance between two points, the angles a 1, a 2 and torsion angle w. w a1a1 a2a2 A B C
Reference Frame Complementarity Budda Knob matches hole for one point only, not necessary for neighbor points. The difference between distances (AB,AC,BC) < d_thr. The difference between the angles < angle_thr. Complexity: O(m 3 n 3 ) Kely Knob matches hole for every point. The difference between distances < d_thr. The difference between the angles < angle_thr. Complexity: O(m 2 n 2 ) w a1a1 a2a2 A B C
Clustering Budda At first fast clustering by transformation parameters is performed. For final results, the RMSD clustering on the interface is applied. Kely The transformations are clustered by the rotational angle between every two conformers.
Geometric Scoring Budda A number of ranges on the surface of receptor molecule is defined: 1.5 For each interface number of surface (shuo) points in each range is computed. The score is a weigted function of each layer. In addition, number of penetrating residues is computed and added to the score. Kely Only 3 ranges are used: Surface, Exterior, Core. The score is given as: GS=S-4E-C 2 /(1+e 6-C ) Additional geom. score: NS = GS/S Electrostatic “clashes” filter Connected components of interface filter.
Capri1: enzyme – inhibitor HPr kinase/phosphatase is a key regulatory enzyme controlling carbon metabolism in bacteria. The protein is a hexamer. HprK/P contains the Walker motif - characterisctic of nucleotide-binding proteins
Capri1: enzyme – inhibitor
It catalyses the ATP-dependent phosphorelation/dephosphorelation of Ser46 in HPr.
Capri1: enzyme – inhibitor The docking was done using Walker motif and Ser46 as active site, so that only relevant solutions could be created. The filter on the distance between phosphate and oxygen atoms of Asp was applied. Only relevant solutions were selected.
Capri 2 and Capri3- biology background Viral capsid:
biology background … Capri2 - VP6 protein of rotavirus that causes gastroenteritis in children. Capri3 – influenza hemagglutinin.
Capri 2
Capri2 antigen... Trimmer. (symmetry) The surface of the B (helices) domain is buried in rotavirus capsid. The H-domain interacts with the antibody. A ‘hint’ was given- to use the trimmer in the docking, meaning that active site is expanded to more than one chain.
Capri 3
Capri3 antigen... Hexamer. 3 dimmers (symmetry) One chain of the dimmer(s) is buried in the capsid Other antibody-antigen complexes of this antigen also implies active site is in the ‘external’ chains (B,D and F)
Antibodies - CDRs The antibody active site is always in the CDRs, which are in the variable part of the antibody CDRs H3 and L3 are most likely to be in the active site. At least 4 CDRs in the active site
CDRs location utility The light and heavy chain has conserved areas which enables us to align a given sequence to a consensus sequence which was built using statistical data. This alignment is then used to locate the CDRS area. This information may be used in docking programs, in the docking process itself (budda) or in a filtering process (kely).
Antibodies Tyr, Trp and Arg propensities Antibodies active sites have a high propensities of Tyr,Trp and Arg. This property may be used to filter out docking results or even be used in the scoring function. (we used both)
Symmetry Both antigens (and also the capri1 enzyme) have a rotational symmetry. ‘rmsd-like distance function’ Clusterring is used to gather transformations which are symmetrically similar.
Clustering different alg results Any algorithms performs well on some targets and less on other. Since 10 results should have been submitted to each target it seemed reasonable to gather results from different algorithms, re-score them, and then re-cluster them.
CAPRI Docking Flow Molecular Surface representation Critical points selection Matching of critical points Clustering and Scoring Compute Connolly and Shuo surface representation. Compute CDRs in case of Antibody- Antigen docking. Select knobs and holes for CDRs only Match critical points - use 2 points and their normals for faster matching. Cluster transforms. Compute geometric score. Apply CDRs, Tyr-Trp, NS, interface connectivity, symmetry filters.
Enzyme – inhibitor test set results (unbound)
Antibody-Antigen test set results (unbound/bound)
New classes and utilities MoleculeGrid class - new class in GAMB++ library. The class can compute molecule grid with distance function, volume function and normals for critical points. CDR utility - utility that finds CDRs for antibody, by aligning new sequence to consensus sequence. Symmetry and Clustering utility New docking version of budda - including new grid representation, two points and normals matching.