Gordon2003
- 1. Designing Nuclear Receptor Targeted Arrays July 10, 2003
- 2. Nuclear Receptor Gene Family Steroid Receptors ER ( , ) estradiol AR dihydrotestosterone PR progesterone MR aldosterone GR cortisol VDR calcitriol LXR ( , ) oxysterols SF1 ( , ) oxysterols CAR androstanes PXR pregnanes FXR bile acids Genome C. elegans D. melanogaster H. sapiens No. of NR Genes 270 21 48 Size (Mb) 80 137 3300 C N DBD LBD Orphan Receptors HNF4 ( , ) NGFIB ( , ) TR2 ( , ) COUP ( , ) ROR ( , ) RevErb ( , ) Tlx ERR ( , ) GCNF DAX SHP LRH TR ( , ) triiodothyronine RAR ( , ) retinoic acid RXR ( , ) 9-cis retinoic acid PPAR ( , ) fatty acids/eicosanoids Non-steroid Receptors
- 3. Nuclear Receptor Signaling Nucleus Cytoplasm Nuclear Receptor DNA +/- cortisol aldosterone progesterone testosterone estradiol thyroid hormone vitamin D retinoic acid
- 4. Lead Discovery Strategies Low High Knowledge Base Random Screening Targeted Screening Database Mining High Low Compound Diversity Array Synthesis Diverse Arrays Targeted Arrays
- 5. NR Chemistry Strategy Array Design molecular descriptors “ NR space” Array Synthesis NR targeted arrays NR compound sets Array Screening screening data virtual screening combichem
- 6. NR Array Design Universe of Compounds (Virtual or Real) Nuclear Receptor Ligands Chemistries Universe of Compounds (Virtual or Real) NR Ligands Chemistries
- 7. Defining NR Chemical Space DiverseSolutions (DVS) Select Descriptors for a Descriptor Space such that: 1) Maximize dimensionality 2) Minimize axes correlation 3) Separate WDI and NR900 NR900 WDI (42,608 cmpds) WDI (42,608 cmpds) NR900 Apply Basis Set of Descriptors 52 Standard 2D and 3D BCUT Metrics SAVOL Molecular Volume 5 Descriptors measure: 1) Charge 2) Polarizability 3) Molecular Shape 4) Molecular Size
- 8. NR Chemical Space NR Chemical Space Training set 900 NR ligands 50 pseudo-3D descriptors Liquid Stores (600k compounds) 70k compounds (E = 12%) Virtual Screen Size Shape (2) Charge Polarizability E = % of compounds in NR space GSK Data World Drug Index Literature
- 9. Measures of Array Profiles 0% Virtual Effectiveness for Target Space Poor Coverage in Target Space ~40% Virtual Effectiveness for Target Space Good Coverage in Target Space Collection A Target Space Target Space Descriptor Axis 1 Descriptor Axis 2 Collection B Target Space Target Space Descriptor Axis 1 Descriptor Axis 2
- 10. Secondary Amide Array 80 Amines X 80 Acids = 6400 Compounds 67% NR Effective
- 13. Virtual Screening 85 Virtual arrays were assembled and analyzed using BCUT descriptors. 10 of these were selected for production
- 19. Resultant arrays screened against 6 nuclear receptors (4 orphans, 2 ligands known). Screening results compared for diverse and targeted sets of compounds. Comparative Screening
- 20. Diverse set of 24,000 compounds (12% E) Targeted set of 8,000 compounds (100% E) 10 Targeted arrays, 18,000 compounds (50% E Avge) Comparative Screening Results
- 21. NR-LBD Phylogeny Ligands Known Ligands Unknown
- 22. Summary BCUT defined NR space has limited utility for Nuclear Receptor array design. Works well for receptor classes with known ligands. Does not work well for orphan receptors as representative ligands are not included in training set.
- 23. NR LBD X-ray Crystallography Steroid/Retinoid 1. AR 2. ER 3. GR 4. MR 5. PR 6. RAR 7. TR 8. VDR Orphans w. Ligands 1. RXR 2. PPAR 3. LXR 4. PXR 5. FXR 6. CAR 7. ERR 8. ROR Orphans w/o. Ligands 1. COUP 2. DAX 3. GCNF 4. HNF4 5. NGFI-B 6. PNR 7. RevErbA 8. LRH 9. SHP 10. SF1 11. TLX 12. TR2 X-ray structure coordinates available: yes no
- 24. Steroid Receptor Focus Large amount of structural data available Established disease association Established high value markets GR for inflammation/asthma AR for androgen deficiency PR for endometriosis and uterine fibroids MR for hypertension/chronic heart failure Goal--find potent, selective, non-steroid ligands
- 25. Array Design Process Generate array chemistry ideas. Enumerate virtual array Use x-ray crystallography data to assess potential ligands by docking virtual arrays into target structures.
- 26. Array Design Process Generate array chemistry ideas. New chemistries, chemistries from literature Analogs of known ligands Compound collections testosterone progesterone cortisol aldosterone
- 27. Array Design Process Enumerate virtual array commercially available and specialized monomers virtual chemistry performed on UNIX machines using CombiLibmaker (arrays up to 20 million) smaller arrays (up to 10,000) can be enumerated using ADEPT (web-based tool)
- 28. Array Design Process Use x-ray crystallography data to assess potential ligands by docking virtual arrays into target structures. Problem can be simplified by assuming that a novel ligand will adopt a similar shape to the ligand crystallized in the LBD. Software is available for comparing shapes and pharmacophores R apid O verlay of C hemical S tructures (ROCS)
- 29. R apid O verlay of C hemical S tructures Performs large scale 3D database searches by using a superposition method that finds the similar but non-intuitive compounds Molecules are aligned by a solid-body optimization process that maximizes the overlap volume between them Volume overlap is not the hard-sphere overlap volume, but rather a gaussian based overlap parameterized to reproduce hard-sphere volumes ROCS for Shape Similiarity
- 30. ROCS for Shape Similarity Shape Match ROCS Crystallized Ligand Virtual Array Shape Score
- 31. O ptimized M olecular E nsemble G eneration A pplication A torsion-driving beam search for generating ensembles of conformers dissects each molecule into smaller fragments by breaking each rotatable bond fragments are reassembled and all of the possible conformations of the subunit being added are sampled Very rapid systematic conformer search OMEGA for Conformations
- 32. OMEGA for Conformations OMEGA Ensemble of Conformers
- 33. Workflow for ROCS Analysis of AR Ligands Virtual Array (SMILES) DHT Shape Query CONCORD for 3D coords OMEGA for Conformers Shape Score Shape Match ROCS for Shape Match
- 34. ROCS Alignment of AR Antagonists with DHT
- 35. Strategy for Virtual High-throughput Docking Large Virtual Array ... Virtual Compounds to be Docked Virtual Compounds to be Synthesized M1 Monomers M2 Monomers M3 Monomers Mx Monomers Enumeration Enumeration Pharmacophore Analysis ROCS Analysis Generation of Conformer Ensemble Dock into Protein Active Site Descriptor Evaluation and Virtual Screening Selection
- 36. Array A Array B
- 37. Array Design Progress 286 Virtual arrays generated Commercial compound collections, in-house ideas, expanded literature chemistries 250 of these have been analyzed 30 High priority arrays have been identified
- 38. Limitations One structure used per protein target Shape match is a rigid body calculation; need to account for “flexibility” of the ligand Current measure of shape similarity is not specific Docking is carried out with a rigid receptor; NRs are known to be quite flexible Analysis throughput limited by computational constraints
- 39. Summary Structure-guided virtual high-throughput screening and compound evaluation are key components to our strategy for pursuing nuclear receptors. A heuristically based virtual docking scheme is in place for library design. Structure is a key component in hit generation. Shown how we generate, analyze, and prioritize virtual arrays using this procedure.
- 40. Acknowledgements Pete Kitrinos Bob Johnson Dean Phelps Melissa Gomez Matt Lochansky Christina Sheedy David Gray Rosemary Sasse Matilde Caivano Vicky Strzelczyk Margaret Clackers Jennifer Brown Ryan Trump Phil Turnbull J.B. Blanc Graham Robinett Peter Brown Bob Wiethe Bill Stuart David Drewry David Jones Andy Noe Bill Hoekstra Tim Willson Frank Schoenen Dudley Rose Donald Lyerly James Ballinger Joyce Turner Brenda Ray Eugene Stewart Mill Lambert Aaron Miller