Looking Back at Mycobacterium tuberculosis Mouse Efficacy Testing To Move New Drugs Forward

Transcription

1 Looking Back at Mycobacterium tuberculosis Mouse Efficacy Testing To Move New Drugs Forward Sean Ekins 1,2, Robert C. Reynolds 3, Antony J. Williams 4, Alex M. Clark 5 and Joel S. Freundlich 6 1 Collaborative Drug Discovery 2 Collaborations in Chemistry 3 University of Alabama at Birmingham 4 Royal Society of Chemistry 5 Molecular Materials Informatics 6 Rutgers University New Jersey Medical School OBJECTS IN MIRROR ARE CLOSER THAN THEY APPEAR

2 Tuberculosis kills m/yr (~1 every 8 seconds) 1/3 rd of worlds population infected!!!!

3 Multi drug resistance in 4.3% of cases Extensively drug resistant increasing incidence one new drug (bedaquiline) in 40 yrs streptomycin (1943) para-aminosalicyclic acid (1949) isoniazid (1952) pyrazinamide (1954) cycloserine (1955) ethambutol (1962) rifampicin (1967)

4 Ponder et al., Pharm Res 31: , 2014

5 3 x published prospective tests >20% hit rate Multiple retrospective tests 3-10 fold enrichment Over 5 years analyzed in vitro data and built models High-throughput phenotypic Mtb screening Mtb screening molecule database/s S Bayesian Machine Learning classification Mtb Model Descriptors + Bioactivity (+Cytotoxicity) N H N Molecule Database (e.g. GSK malaria actives) virtually scored using Bayesian Models Top scoring molecules Ekins et al., Pharm Res 31: , 2014 assayed for Ekins, et al., Tuberculosis 94; , 2014 Mtb growth inhibition Ekins, et al., PLOSONE 8; e63240, 2013 Ekins, et al., Chem Biol 20: , 2013 Ekins, et al., JCIM, 53: , 2013 Ekins and Freundlich, Pharm Res, 28, , 2011 Ekins et al., Mol BioSyst, 6: , 2010 Ekins, et al., Mol. Biosyst. 6, , 2010, Identify in vitro hits and test models New bioactivity data may enhance models

6 Taking a compound in vivo identifies issues BAS / TCMDC reported to be a P. falciparum lactate dehydrogenase inhibitor Only one report of antitubercular activity from solid agar MIC = 1 mg/ml ( wild strain ) - no activity in mouse model up to 400 mg/kg - however, activity was solely judged by extension of survival! MIC of ug/ml 64X MIC affords 6 logs of kill Resistance and/or drug instability beyond 14 d Vero cells : CC 50 = 4.0 mg/ml Bruhin, H. et al., J. Pharm. Pharmac. 1969, 21, Selectivity Index SI = CC 50 /MIC Mtb = In mouse no toxicity but also no efficacy in GKO model probably metabolized. Ekins et al.,chem Biol 20, , 2013

7 Big Data: Screening for New Tuberculosis Treatments Tested >350,000 molecules Tested ~2M 2M >300,000 >1500 active and non toxic Published s 800 Others have likely screened another 500,000 How many will become a new drug? How do we learn from this big data?

8 Can combining Big and Small data (in vitro, in vivo) help us find better compounds, faster? Avoid testing as many molecules Need to ensure PK/ADME properties are good

9 What is the next bottleneck? $ $ $ $ $ Five-fold increase in the publication of TB mouse model studies from 1997 to 2009 Franco, PLoS One 7, e47723 (2012).

10 Billions of $ of your money spent on TB but no database of mouse in vivo data!

11 Hunting for the in vivo data It s out there.. be patient

12 Data sources PubMed, SciFinder (CAS, Columbus OH) Web of Knowledge (Thomson Reuters) Individual journals (e.g. Tuberculosis, the Journal of Medicinal Chemistry, and PLOS journals). Searched for papers with compounds tested in murine acute and chronic Mtb infection models. tuberculosis and in vivo and mouse, Tuberculosis and efficacy and mouse and comparison and antituberculosis and mouse. Compounds from >100 sources -papers, books, slides

13 Building the mouse TB database Manually curated, structures sketched Mobile Molecular DataSheet (MMDS) ios app or ChemDraw (Perkin Elmer) Downloaded from Combined with pertinent data fields 1 log 10 reduction in Mtb colony-forming units (CFUs) in the lungs Publically available CDD TB database (In process)

14 Mean of molecular descriptors for N = 773 in vivo Mtb dataset, comparing actives and inactives Num Arom MW AlogP HBD HBA Num Rings Rings FPSA RBN Active (N = 362) ± ± 2.71* 1.49 ± ± ± 2.09* Num 1.72 Arom ± ± 0.13* 7.84 ± Inactive Active MW AlogP HBD HBA Num Rings Rings FPSA RBN (N = 411) ± ± ± ± ± ± ± ± (N = 362) ± ± 2.71* 1.49 ± ± ± 2.09* 1.72 ± ± 0.13* 7.84 ± Inactive The AlogP observation is opposite to what we see for in vitro data (N = 411) ± ± ± ± ± ± ± ± 16.05

15 30 MIND years with THE little TB GAP mouse in vivo data

16

17 Where are the New TB drugs to be found? PCA of In vivo actives and inactives Inactive (blue) Active (yellow) PCA of TB in vitro actives (blue) TB in vivo (yellow) PCA of in vivo (yellow) and compounds with known targets (blue)

18 Machine Learning Models Bayesian Support Vector Machine Recursive partitioning (single and multiple trees) Using Accelrys Discovery Studio and R. RP Forest RP Single SVM Bayesian Tree ROC 5 fold cross validation

19 Bayesian Model Studio and R. Leave-one- Leave-out Leave-out Leave-out 50% Leave-out Leave-out 50% out ROC 50% x % x 100 x % x 100 x 100 External ROC Internal Concordance Specificity Sensitivity Score ROC Score ± ± ± ± ± 9.19 Model statistics are reasonable

20 External Test set Studio and R. 11 additional active molecules obtained from RP Forest RP Single SVM Bayesian Tree 3 /11 4/11 7/11 8/11 (27.2%) (36.4%) (63.6%) (72.7%)

21 Bayesian Models: Back to the Dark Ages 24 consensus actives in vivo computational models

22 MoDELS RESIDE IN PAPERS NOT ACCESSIBLE THIS IS UNDESIRABLE How do we share them? How do we use Them?

23 ECFP_6 FCFP_6 Collected, deduplicated, hashed Sparse integers Invented for Pipeline Pilot: public method, proprietary details Often used with Bayesian models: many published papers Built a new implementation: open source, Java, CDK stable: fingerprints don't change with each new toolkit release well defined: easy to document precise steps easy to port: already migrated to ios (Objective-C) for TB Mobile app Provides core basis feature for CDD open source model service

24 Same datasets Versus published data Dataset Leave one out ROC Published In vivo data (773 molecules) FCFP_6 fingerprints Combined model (5304 molecules) FCFP_6 fingerprints MLSMR dual event model (2273 molecules) and FCFP_6 fingerprints Reference Leave one out ROC Open fingerprints 0.77 this study J Chem Inf Model 53: PLOSONE 8:e

25 Open fingerprints and bayesian method used in TB Mobile Vers.2 Predict targets Cluster molecules Could we add in vivo prediction models to this? Ekins et al., J Cheminform 5:13, 2013 Clark et al., submitted 2014

26 Optimal Mouse properties Optimal TB entry properties Optimal Human properties

27 In vitro data In vivo data Target data ADME/Tox data & Models Data sources and tools we could integrate Drug-like scaffold creation TB Prediction Tools TB Publications

28 We could do this again for a different disease Small data: Mouse In vivo model data «Tuberculosis and mouse in vivo model» ~338 papers in PubMed «Malaria and mouse in vivo model» ~314 papers in PubMed

29 The Clock is ticking A much higher ratio of compounds were tested in vivo to in vitro in the 1940s-1960s rather than now Infrastructure to provide a clear understanding of the position of compounds in the pipeline is essentially lacking Shortage of new candidates suggest we may lack the commitment and resources we had 6o years ago Use machine learning in vivo models to prioritize Mouse studies Ekins, Nuermberger & Freundlich Submitted

30 Replacement Reduction Refinement Literature data on in vivo testing ~ 1500 molecules Descriptors + Bioactivity Bayesian, SVM, Trees etc Test models with in vivo testing after in vitro ADME and in vivo PK

31 TB Mobile Vers 2 Is on itunes and Google play and it is FREE

32 Acknowledgments: Richard S. Pottorf, Anna Coulon Spektor, Dr. Barbara Laughon 2R42AI from the National Institute of Allergy And Infectious Diseases. (PI: S. Ekins) R43 LM from the National Library of Medicine (PI S. Ekins).