Accelrys UGM slides 2011

Using Discovery Studio For Modeling Human Drug Transporters Sean Ekins Collaborations in Chemistry, Jenkintown, PA. School of Pharmacy, Department of Pharmaceutical Sciences, University of Maryland.

Transporter models in ADME Absorption Distribution Excretion Metabolism Chang, C. and Swaan, P. in Ekins S, Computer Applications in Pharmaceutical Research and Development, pp495-512, 2006 PEPT1 ASBT OCTN2 NT MDR (P-gp, MRP, BCRP) OATP OAT OCT MCT NTCP BSEP BBB CHT PEPT1

Nature Reviews Drug Discovery 9 , 215–236 (1 March 2010)

Why We Need Transporter Models Predict if a compound will interact at a transporter Potential for drug – drug or drug – endogenous compound interactions Impact bioavailability or result in Toxicity Start understanding ligand-protein interactions in vitro models - limited throughput & in vivo study complicated by multiple transporters with overlapping substrate specificities. in silico – in vitro approach has value in targeting testing of compounds with a high probability of activity. New industry FDA recommendations include transporters Nature Reviews Drug Discovery 9 , 215–236 (1 March 2010)

From one extreme to another A decade ago focus was P-gp and we had small datasets Now we may have massive (tens of thousands of compounds) datasets in pharma Over time - Bigger datasets, more complex models + use of more methods Shift in interest to other transporters – lots of them - datasets still very small or non existent Predominant - inhibitor data

Ideal when we have few molecules for training In silico database searching Accelrys Catalyst in Discovery Studio Geometric arrangement of functional groups necessary for a biological response Generate 3D conformations Align molecules Select features contributing to activity Regress hypothesis Evaluate with new molecules Excluded volumes – relate to inactive molecules Pharmacophores applied broadly Created for CYP2B6 CYP2C9 CYP2D6 CYP3A4 CYP3A5 CYP3A7 hERG P-gp OATPs OCT1 OCT2 BCRP hOCTN2 ASBT hPEPT1 hPEPT2 FXR LXR CAR PXR etc

Pharmacophore Models Substrate Model 1 (aligned with verapamil) The Catalyst HIPHOP model is generated based on verapamil and digoxin. 5 hydrophobic (HYD) features (cyan) 2 hydrogen bond acceptor (HBA) features (green) Inhibitor Model 1 (aligned with LY335979) The model is calculated based on 27 diverse compounds with IC 50 for inhibition of [ 3 H]-digoxin transport in Caco-2. 4 HYD features (cyan) 1 HBA Feature (green) Inhibitor Model 2 (aligned with CP114416) The model is an update of above model with six additional inhibitors (FK506A, PSC 833, Ritonavir, Erythromycin, Cyclosporin and CP99542). 4 HYD features (cyan) 1 HBA feature (green) Chang, Bahadurri et al, DMD 34, 1976-1984 (2006)

Pharmacophore Development Database screening 189 known P-gp substrates and non-substrates 576 prescription drugs. In silico validation In vitro validation Substrate Model Inhibitor Model 1 Inhibitor Model 2 Efflux Ratio ATPase Activation assay Güner-Henry Score 33 compounds inhibition of [3H]-digoxin transport in Caco-2) Chang, Bahadurri et al, DMD 34, 1976-1984 (2006) A B

Summary of P-gp prospective screen Evolution and application of earlier pharmacophores Pharmacophores retrieve ~50% known P-gp substrates or inhibitors in hit lists from database searching Discovered novel P-gp interacting classes Potassium channel blocker (Repaglinide) Retinoid (Acitretin) Nonpeptide angiotensin II antagonist (Telmisartan) Prostaglandin analogue (Misoprostal) Expanded P-gp interaction profile Penicillin based  -lactam antibiotic (Nafcillin) Steroid hormone (cholecalciferol)  2 -adrenoceptor agonist (Salmeterol) Chang, Bahadurri et al, DMD 34, 1976-1984 (2006)

MRP BCRP P-gp Molecule Databases In vitro testing hPEPT Transporter Pharmacophores or other model types Feedback of new substrates or inhibitors More Transporters - More Models Ekins, in Ecker G and Chiba P, Transporters as drug carriers, John Wiley and Sons. P215-227, 2009. MRP BCRP P-gp Molecule Databases In vitro testing hPEPT Transporter Pharmacophores Feedback of new substrates or inhibitors

95% 5% >95% 5% Bile acid pool: 3-5g Circulate 6-10 daily Total turnover rate: 20-30g Lost in feces: < 0.5g Enterohepatic Circulation Expressed at high levels in the terminal ileum where it mediates bile acid recovery. studies have implicated secondary bile acids as important in the development of colorectal cancer. hASBT inhibition results in increased colonic exposure to cytotoxic secondary bile acids. an association between a polymorphism in the SLC10A2 gene and the risk of colorectal adenomatous polyps. The human Apical Sodium-dependent Bile Acid Transporter Zheng X, et al., Mol Pharm, 6: 1591-1603, 2009

Computational Models for ASBT Common feature (HipHop) Pharmacophore Modeling Identify common molecular features based on the superposition of active compounds Arrange key features that are important for biological activity Predicts the compound as “active” or “inactive” based on fit to the pharmacophore Quantitative Pharmacophore Modeling A quantitative pharmacophore was generated using the HypoGen method. The statistical relevance of the pharmacophore was evaluated based on structure activity correlation coefficient and cost value relative to the null hypothesis. Bayesian Model Classification model uses fragment and simple interpretable descriptors Aid in understanding features in actives and inactives SCUT Database Total 796 compounds (656 drugs in the SCUT 2008 database plus additional drug metabolites and drugs of abuse)

Computational Models for ASBT - process 38 Bayesian models + validation with test sets Test set 1 N= 30 from same lab Test set 2 N = 19 from literature sources

HipHop Pharmacophore Model A common feature HipHop pharmacophore model of hASBT was generated from the 11 most potent inhibitors. The most active compound mesoridazine was used to create a shape restriction. retrieved 58 hits - 15 selected and assessed for ASBT inhibition. 8 were were potent inhibitors ( K i < 100 μ M). The model consisted of: 2 hydrogen bond acceptors (Green) 2 hydrophobes (Blue) Zheng X, et al., Mol Pharm, 6: 1591-1603, 2009

Calcium Channel Blockers and HMG CoA-reductase Inhibitors Most potent inhibitors were calcium channel blockers and HMG CoA-reductase inhibitors Five from secondary screen (*) Additional calcium channel blockers and statins were studied. Dihydropyridines are more potent hASBT inhibitors than other classes of calcium channel blockers . Zheng X, et al., Mol Pharm, 6: 1591-1603, 2009 Compound Drug Class K i Value ( µM) Nifedipine Dihydropyridine 3.87±0.64 Nisoldipine Dihydropyridine 4.77 ± 1.0 5 Nimodipine* Dihydropyridine 5.75±0.72 Simvastatin* Statin 10.4± 2.1 Fluvastatin* Statin 11.5±0.8 Isradipine Dihydropyridine 19.4±3.0 Lovastatin* Statin 21.6±2.3 Nemadipine Dihydropyridine 23.1±4.1 Nicardipine Dihydropyridine 32 . 4 ± 3 . 1 Nitrendipine Dihydropyridine 34.1± 5.1 Amlodipine* Dihydropyridine 42.1±7.7 Felodipine Dihydropyridine 49.7 ± 7.0 Diltiazem Benzothiazepines 211 ± 21 Verapamil Phenylalkylamine 26 6 ± 2 2

Quantitative Pharmacophore Model hASBT A quantitative pharmacophore was generated with 38 compounds using the HypoGen method The model consisted of: 1 hydrogen bond acceptors (Green) 3 hydrophobes (Blue) 5 excluded volumes r : 0.815. Zheng X, et al., Mol Pharm, 6: 1591-1603, 2009

Bayesian machine learning Ekins, Williams and Xu, Drug Metab Dispos 38: 2302-2308, 2010 Bayesian classification is a simple probabilistic classification model. It is based on Bayes’ theorem h is the hypothesis or model d is the observed data p ( h ) is the prior belief (probability of hypothesis h before observing any data) p ( d ) is the data evidence (marginal probability of the data) p ( d|h ) is the likelihood (probability of data d if hypothesis h is true) p ( h|d ) is the posterior probability (probability of hypothesis h being true given the observed data d ) A weight is calculated for each feature using a Laplacian-adjusted probability estimate to account for the different sampling frequencies of different features. The weights are summed to provide a probability estimate

Molecular function class fingerprints of maximum diameter 6 (FCFP_6), + simple interpretable descriptors leaving 20% out 100 times, ROC was 0.78; concordance 72.5%; specificity 81.0%; Sensitivity 58.1%. Bayesian and pharmacophore perform similarly > 80% correct with n= 30 test set All models perform poorly with literature test set Bayesian Model hASBT +ve -ve Dihydropyridine substructure Zheng X, et al., Mol Pharm, 6: 1591-1603, 2009

ASBT Conclusions A large number of drugs from diverse classes were found to be hASBT inhibitors. Most potent inhibitors were dihydropyridine calcium channel blockers, HMG CoA-reductase inhibitors, or diuretics. A HipHop pharmacophore and a quantitative pharmacophore of hASBT was generated and have features in common. Literature test set – poor predictions with both models – variability in data Future pharmacoepidemiologic studies will attempt to elucidate a possible relationship between hASBT inhibition and the incidence of colon cancer. Current data conflicting on role. Novel insights from in vitro and in silico methods for ASBT interactions Zheng X, et al., Mol Pharm, 6: 1591-1603, 2009

hOCTN2 High affinity cation/carnitine transporter - expressed in kidney, skeletal muscle, heart, placenta and small intestine Inhibition correlation with muscle weakness - rhabdomyolysis A common features pharmacophore developed with 7 inhibitors Searched a database of over 600 FDA approved drugs - selected drugs for in vitro testing. 33 tested drugs predicted to map to the pharmacophore, 27 inhibited hOCTN2 in vitro Compounds were more likely to cause rhabdomyolysis if the C max / K i ratio was higher than 0.0025 Diao, Ekins, and Polli, Pharm Res, 26, 1890, (2009)

Possible Association between Clinical Rhabdomyolysis and hOCTN2 Inhibition Diao, Ekins, and Polli, Pharm Res, 26, 1890, (2009)

+ve -ve hOCTN2 quantitative pharmacophore and Bayesian model Diao et al., Mol Pharm, 7: 2120-2131, 2010 r = 0.89 vinblastine cetirizine emetine

hOCTN2 quantitative pharmacophore and Bayesian model Bayesian Model - Leaving 50% out 97 times external ROC 0.90 internal ROC 0.79 concordance 73.4%; specificity 88.2%; sensitivity 64.2%. Lab test set (N = 27) Bayesian model has better correct predictions (> 80%) and lower false positives and negatives than pharmacophore (> 70%) Predictions for literature test set (N=32) not as good as in house – mean max Tanimoto similarity were ~ 0.6 Diao et al., Mol Pharm, 7: 2120-2131, 2010 PCA used to assess training and test set overlap

Among the 21 drugs associated with rhabdomyolysis or carnitine deficiency, 14 (66.7%) provided a C max/ K i ratio higher than 0.0025. Among 25 drugs that were not associated with rhabdomyolysis or carnitine deficiency, only 9 (36.0%) showed a C max / K i ratio higher than 0.0025. Rhabdomyolysis or carnitine deficiency was associated with a C max / K i value above 0.0025 (Pearson’s chi-square test p = 0.0382). limitations of C max / K i serving as a predictor for rhabdomyolysis -- C max / K i does not consider the effects of drug tissue distribution or plasma protein binding. hOCTN2 association with rhabdomyolysis Diao et al., Mol Pharm, 7: 2120-2131, 2010

hOCTN2 Substrates Data from Polli lab (conjugates) and literature 161 ± 50 Valproyl-glycolic acid-L-carnitine 58.5 ± 8.7 Ketoprofen-glycine-L-carnitine 77.0 ± 4.0 Ketoprofen-L-carnitine 257 ± 57 Naproxen-L-carnitine 132 ± 23 Valproyl-L-carnitine 53 Ipratropium 26 Mildronate 9 Acetyl-L-carnitine 5.3 L-carnitine Km (microM) Substrate

Substrate Common feature Pharmacophore ---Used CAESAR and excluded volumes Inhibitor Hypogen pharmacophore Overlap of pharmacophores RMSD 0.27 Angstroms hOCTN2 Pharmacophores

Proactive database searching - Prioritize compounds for testing in vitro In silico allows rapid parallel optimization vs transporters or other properties Partial overlap of OCTN2 of substrate and inhibitor pharmacophores Work continuing on 3 additional transporters with academic collaborators Discovery Studio Pharmacophore and Bayesian components enable fast model building and database searching – in vitro data generation is rate limiting step Provide novel insights into the molecular interactions with transporters Summing up

Future … Virtually every type of algorithm/ descriptor combination used on available P-gp data – this will happen to others Need for much larger public / available datasets – These will likely be used with machine learning algorithms Transporter databases in the future with on line models? Limited accessibility of models Applicability domain of models / confidence in predictions rarely evaluated Use experience to model other transporters, find substrates etc Chang, Bahadurri et al, DMD 34, 1976-1984 (2006)

Open source tools for transporter modeling Open source descriptors CDK and C5.0 algorithm ~60,000 molecules with P-gp efflux data from Pfizer MDR <2.5 (low risk) (N = 14,175) MDR > 2.5 (high risk) (N = 10,820) Test set MDR <2.5 (N = 10,441) > 2.5 (N = 7972) Could facilitate model sharing? Gupta RR, et al., Drug Metab Dispos, 38: 2083-2090, 2010 $ $$$$$$

2D Similarity search with “hit” from transporter screening Export database and use for 3D searching with a pharmacophore or other model for transporter Suggest approved drugs for testing - may also indicate other uses if it is present in more than one database Suggest in silico hits for in vitro screening Key databases of structures and bioactivity data FDA drugs database Could future transporter models help Repurpose FDA drugs

Acknowledgments University of Maryland Xiaowan Zheng Lei Diao Cheng Chang Praveen Bahadduri Peter W. Swaan James E. Polli Pfizer Rishi Gupta Eric Gifford Ted Liston Chris Waller Accelrys [email_address] http://www.slideshare.net/ekinssean http://www.collaborations.com/CHEMISTRY.HTM Twitter: collabchem Blog: www.collabchem.com

Accelrys UGM slides 2011

More Related Content

What's hot (20)

Similar to Accelrys UGM slides 2011 (20)

More from Sean Ekins (20)

Recently uploaded (20)

Accelrys UGM slides 2011

Editor's Notes