$NIGMS $FDA The Central Dogma of Molecular Pharmacology (~1985 to - PowerPoint PPT Presentation

Bryan Roth Brian Kobilka, Peter Gmeiner Matt O’Meara, Josh Pottel Henry Lin, Anat Levit Kate Stafford Magdalena Korzynska John Irwin $NIGMS $FDA

The Central Dogma of Molecular Pharmacology (~1985 to present): Target à Ligand

Docking for new chemotypes from large libraries 10 5 complexes/molecule 10 7 available compounds (~10 12 complexes overall) Test high-scoring molecules (but which ones, exactly?)

Docking for novel agonists with new µ OR biology N OH O OH 3.5M cmpds Test S O O N N N H H N N N H H HO N Cl K i 1 nM, 7/23 tested hit, 2 to 14 uM EC 50 4 nM Manglik, Lin, Aryal… Nature 2016

docking screens vs GPCRs: 17 to 58% hit rates, nanomolar activities, novel ligands docking x-ray b 2AR: 25%, muscarinic: 58% A2a: 35%, D3 model: 23% 0.01 to 3 uM 0.4 to 40uM 0.2 to 3uM Kolb, PNAS 2009 0.2 to 3 uM 3x10 6 Kruze, Mol Pharm 2013 Carlsson, Nat Chem Carlsson, JMC 2010 Biol 2011 cmpds Test a probe for the a probe for the CXCR4:17%, orphan MRGPRX2 orphan GPR68 0.3 to 30uM Lansu, unpub Haung, Nature 2015 Mysinger, PNAS 2012

(some) inaccuracies in docking D G interact - D G solv, L - D G solv, R = D G bind = D G interac t S (q i P i + v i P v ) ((1/D 0 - 1/D w )/2r S Q 2i - Born)*a i d V/ S 1/ r 4ik + Δ H np + IST = D G solv Lennard-Jones Potential PB electrostatics grid 20.000 A B = - f ( x ) Energy (kcal/mol) 12 6 r r X = 1.50Å O N + 0.00 0.50 1.00 1.50 2.00 2.50 3.00 -5.000 NH Radius (Å) Charges & params for 10 7 diverse molecules O NH N + internal energies NH O Relaxation O NH O - Hydrophobicity (solvent effects) Water displacement...

The data are sparse, the space is big • <10 5 crystallographic ligand complexes • 10 6 ligand-protein affinities measured • Many badly measured, many artifactual • Over 10 62 possible drug-like molecules • Small differences in chemical structure can matter N + N N N N N N N O N HO N HO N H

Opportunities to collaborate: model systems, late state prioritization, next compound… simplified sites, readily tested at atomic resolution 10 4 (…10 5 … 10 6 ) good docking hits, which to test? Optimize?

The upside down world of classical pharmacology (1930 to 1985: ligands à targets) OH HO a O HO HO > > HN NH NH 2 O HN HO HO OH NH 2 HO adrenergic HO HO HO b 1 adrenergic O HN HO HO HO > > HN OH b adrenergic NH 2 H NH N HO HO HO HO HO HO HO HO b 2 adrenergic

Relate targets by ligand similarity (~10 6 ligands for ~2500 targets)

656 drugs predicted vs 73 side effect targets Statistical model 1 For each of 656 drugs, drug vs. t B Frequency 2 Compare to ligands of 73 Targets drug vs. t A Z-score , … 3 Test at Novartis t A t A t B E. Loukine, Nature 2012

0 to 47888 possible drug-target pairs. 1241 novel ones predicted, 1042 tested at Novartis in proprietary dbs 65 unknown to SEA Confirmed <30 uM 348 Confirmed <10 uM 478 Confirmed < 1 uM Inactive 77 26 48 Ambiguous Lounkine, Nature 2012

Example ADR targets (26% cross domain boundaries) Closest SEA E- Off- Known BLAST Kd Drug known value Target Target E (nM) ADR 1.6e -17 5HT2b 5HT3 N N N sedation N >>1 20 N O NH O F Alosetron Tc 0.25 5HT7 F F O NH N 4.8e- 53 207 N Ventricular O 3e -7 O a 1a N arrhythmia N HN O Pimozide F Tc 0.43 O O Cl OH O Estr. 1.9e -17 COX1 221 N >>1 abdominal O Rec Cl O pain Tc 0.31 O chlorotrianisene Cl Cl hERG >>1 Palpitations, O HO 1.1e -14 SERT 419 N Insomnia N O NH Clemastine Tc 0.32 Loukine, Nature 2012

Reorganize the GPCR-ome by ligand similarity Statistical model A vs. C Frequency A A 1 For each GPCR 2 ligand set A vs. B ? ? Z-score 4 Rank by significance … B C 3 vs. the ligand sets for 3000 targets Lin, Nature Methods , 2013 Keiser, Nature Biotech 2007

GPCRs by orthosteric seq. id GPCRs by ligand similarity Bioamine Peptide non-GPCRs 6 predicted Lipid Purine w/GPCR- crosses tested, Adenosine like ligands confirmed Melatonin Gloriam, JMC 2009 Lin, Nature Methods 2013

A 250 nM casein kinase1 inhibitor predicted to modulate CXCR2 (E-value 1.3 x10 -15 , EC50 780 nM) EC50 = 0.78 uM Lin, Nature Methods , 2013 (Bryan Roth, Flori Sassano)

An epoxide hydrolase inhibitor predicted to modulate CB2 (E-value 1.3 x10 -15 , EC50 2.6 uM) CB2 Inhibition by Epox hyd inhib WIN 55,212-3 Lin, Nature Methods 2013

How general is this?

Ligand similarity reorganizes the genome NHRs by seq. ID…by ligand similarity Matt O’meara Sarah Barelier

Ligand similarity reorganizes the genome: LGICs by seq. ID… by ligand similarity

What is going on?

How unrelated receptors bind identical ligands Sarah Barelier et al, ACS Chem Biol 2015

A “Metabolic Code” : primary messengers fixed, signal in multiple time domains GM Tomkins, Science 1975

GPCR neurotransmitter polypharmacology (the drugs mimic the neurotransmiters) neurotensin Adrenaline 5HT DA ATP adenosine ach sphingosine Lin, O’Meara, leukotriene prostacyclin Barelier, unpub

Nuclear hormone polypharmacology (the drugs mimic the hormones) O’Meara, Barelier, unpub

LGIC neurotransmitter polypharmacology O’Meara, Barelier, unpub

Opportunities : better fingerprints, bioinformatic +chemoinformatic integration, what disease? in proprietary dbs New fingerprints, methods to do 65 unknown to SEA Confirmed <30 uM 348 better than 50% predictive true Confirmed <10 uM positives, 4% predictive false 478 Confirmed < 1 uM negatives Inactive 77 26 48 Ambiguous What polypharmacology exactly should be targeted? O O OH HO N N H H OH HO N N Matt O’Meara