Genomic Dose Response
Total Page:16
File Type:pdf, Size:1020Kb
Genomic Dose Response: The Picture NTP Genomic Dose Response Modeling Expert Panel Meeting October 23, 2017 Russell Thomas Director National Center for Computational Toxicology The views expressed in this presentation are those of the author and do not necessarily reflect the views or policies of the U.S. EPA Scott May Want To Rethink Asking Me To Be The “Big Picture Guy”… National Center for Computational Toxicology It is a Well Known Fact that Toxicology Continues to Have a Data Problem US National Research Council, 1984 Size of Estimate Mean Percent Category Category In the Select Universe Pesticides and Inert Ingredients of Pesticides 3,350 Formulations 10 24 2 26 38 Cosmetic Ingredients 3,410 2 14 10 18 56 Drugs and Excipients Used in Drug Formulations 1,815 18 18 3 36 25 Food Additives 8,627 • Major challenge is too many 5 14 1 34 46 Chemicals in Commerce: chemicals and not enough At Least 1 Million 12,860 Pounds/Year data 11 11 78 Chemicals in Commerce: • Total # chemicals = 65,725 Less than 1 Million 13,911 Pounds/Year • Chemicals with no toxicity 12 12 76 data of any kind = ~46,000 Chemicals in Commerce: Production Unknown or 21,752 Inaccessible 10 8 82 Complete Partial Minimal Some No Toxicity Health Health Toxicity Toxicity Information Hazard Hazard Information Information Available Assessment Assessment Available Available 2 Possible Possible (But Below Minimal) National Center for Computational Toxicology It is a Well Known Fact that Toxicology Continues to Have a Data Problem 70 60 50 40 30 20 Percent of Chemicals <1% 10 0 Acute Cancer Gentox Dev Tox Repro Tox EDSP Tier 1 Modified from Judson et al., EHP 2009 3 National Center for Computational Toxicology For Those With Data, Have We Been Truly Predictive or Just Protective? …data compiled from 150 compounds with 221 human toxicity events reported. The results showed the true positive human toxicity concordance rate of 71% for rodent and nonrodent species, with nonrodents alone being predictive for 63% of human toxicity and rodents alone for 43%. National Center for Computational Toxicology Budding Field of Toxicogenomics Promised to Change All That… Rapid Mode Hazard Of Characterization Action National Center for Computational Toxicology Initial Focus of Toxicogenomics Was on Inferring MOA -Log P-value 0 2 4 6 8 10 1. Neurophysiological process_Circadian rhythm 2. Cytoskeleton remodeling_Cytoskeleton remodeling • Initial 3.a NRF2pplications regulation of oxidative stress response of toxicogenomics to identify MOA have generally 4. Cytoskeleton remodeling_TGF, WNT and cytoskeletal lacked a systematicremodeling approach 5. Development_BMP signaling • More 6.art Cell adhesion_Role than of tetraspanins science in the integrin- mediated cell adhesion • Development7. Signal transduction_JNK pathway of large reference databases difficult due to cost ME Cellular Effects Tissue Effect Organ Effect constraints5 day, 90 ppm 8. Development_EGFR signaling via PIP3 5 day, 200 ppm 9. Cell adhesion_Integrin-mediated cell adhesion and Hepatocyte Cell Liver 15 day, 30 ppm migration ER Stress Cell Death Hyperplasia • Most expert committees/reports defaultedInjury to usingProliferation it as partTumors of an 15 day, 90 ppm overall15 day, 200 ppm 10. Methionine weight-cysteine-glutamate- metabolismof-evidence • Not very satisfying MOA Generator National Center for Computational Toxicology Focus Shifted Towards Supervised Classification Approaches • Most studies show 60-85% accuracy for predicting cancer- related endpoints • Only a limited number of tissues have been evaluated • Requires >20 compounds with adequate redundancy and diversity in mechanisms to have a robust training set (Thomas et al., 2009) • >30 organs show tumor responses in NTP database with ~50% having >10 chemicals in at least one species/sex • Difficult to justify as a comprehensive screen for rodent carcinogenicity National Center for Computational Toxicology Continued Evolution Toward More Quantitative Applications TP53 BMDL NFkB RfD UFs BMR Apical POD Apical BMDL BMD BMDL10 RSD 100,000 Transcriptional POD ShortSubchronic-term Animaland TranscriptionalMost Sensitive FitFit Each Data Genewith IdentifyIdentifyGroup PointPoint Genes ofof By DeriveCompare Safe In Vivo ChronicStudies Animal PathologicalResponses withStatistical Statistical DepartureSignaling (i.e., BMD) Pathway ExposureApical Level and Studies Response Models and Calculate Transcriptional Considered Summary Value PODs Adverse (usually the median pathway BMD) Thomas et al., Tox Sci., 2011 Thomas et al., Mut Res., 2012 Thomas et al., Tox Sci., 2013 National Center for Computational Toxicology In Vivo Study to Assess Transcriptional and Apical Correlation Chemical Route Dosesc Rodent Model Time Point Target Tissue 1,2,4-Tribromobenzenea Gavage 2.5, 5, 10, 25, 75 mg/kg Male Sprague Dawley rats 5 d, 2, 4, 13 wks Liver Bromobenzenea Gavage 25, (50), 100, 200, 300, 400 mg/kg Male F344 rats 5 d, 2, 4, 13 wks Liver 2,3,4,6- Gavage 10, 25, 50, 100, 200 mg/kg Male Sprague Dawley rats 5 d, 2, 4, 13 wks Liver Tetrachlorophenola 4,4'-Methylenebis (N,N- Feed 50, 200, 375, 500, 750 ppm Male F344 rats 5 d, 2, 4, 13 wks Thyroidb dimethyl) benzenamineb N-Nitrosodiphenylamineb Feed 250, 1000, 2000, 3000, 4000 ppm Female F344 rats 5 d, 2, 4, 13 wks Bladderb HydrazobenzeneMeasuredb apical (histologicalFeed 5, and 20, 80, organ 200, 300 weight; ppm n = 10) andMale F344gene rats expression5 d, 2, changes4, 13 wks (nLiver = 5)b at each dose and time point in the target tissue. 1,4-Dichlorobenzeneb Gavage 100, 300, 400, 500, 600 mg/kg Female B6C3F1 mice 13 wks Liver Propylene glycol mono-t- Inhalation 25, 75, 300, 800, 1200 ppm Female B6C3F1 mice 13 wks Liver butyl etherb 1,2,3-Trichloropropaneb Gavage 2, 6, 20, 40, 60 mg/kg Female B6C3F1 mice 13 wks Liver Methylene Chlorideb Inhalation 100, 500, 2000, 3000, 4000 ppm Female B6C3F1 mice 13 wks Liver, Lung Naphthaleneb Inhalation 0.5, 3, 10, 20, 30 ppm Female B6C3F1 mice 13 wks Lung 1,4-Dichlorobenzeneb Gavage 100, 300, 400, 500, 600 mg/kg Female B6C3F1 mice 13 wks Liver aChemicals in IRIS database for non-cancer endpoints only bChemicals previously tested by the U.S. National Toxicology Program cUnderlined doses used in NTP two-year rodent bioassay or IRIS database National Center for Computational Toxicology Temporal Changes Between Transcriptional and Non-Cancer PODs 1000 1000 5 Days 2 Weeks Bladder Liver Thyro id 100 100 10 10 Lowest Apical BMD (mg/kg/d) BMD Apical Lowest Lowest Apical BMD (mg/kg/d) BMD Apical Lowest r = 0.909 r = 0.987 1 1 1 10 100 1000 1 10 100 1000 Lowest Pathway Transcriptional BMD Lowest Pathway Transcriptional BMD (mg/kg/d) (mg/kg/d) 1000 1000 4 Weeks 13 Weeks 100 100 10 10 1 Lowest Apical BMD (mg/kg/d) BMD Apical Lowest Lowest Apical BMD (mg/kg/d) BMD Apical Lowest r = 0.985 r = 0.982 1 0.1 1 10 100 1000 0.1 1 10 100 1000 Lowest Pathway Transcriptional BMD Lowest Pathway Transcriptional BMD (mg/kg/d) (mg/kg/d) Thomas et al., Toxicol Sci, 2013 National Center for Computational Toxicology Combined Correlation Between Non- Cancer and Transcriptional PODs 10000 Bladder Median Log2 Ratio Apical:Transcriptional BMD = 0.068 (1.05) Liver Thyro id 1000 Lung Rat Mouse 100 10 Lowest Apical BMD (mg/kg/d) BMD Apical Lowest 1 r = 0.827 (p = 0.0031) 0.1 0.1 1 10 100 1000 10000 Lowest Pathway Transcriptional BMD (mg/kg/d) Thomas et al., Toxicol Sci, 2013 National Center for Computational Toxicology Temporal Changes Between Transcriptional and Cancer PODs 1000 Bladder 100 Thyroid 10 Liver Tumor BMD (mg/kg/d) BMD Tumor 1 5 d 2 wk 4 wk 13 wk 0.1 0.1 1 10 100 1000 Thomas et al., Toxicol Sci, 2013 Lowest Pathway Transcriptional BMD (mg/kg/d) National Center for Computational Toxicology Combined Correlation Between Cancer and Transcriptional PODs 10000 Median Log2 Ratio Apical:Transcriptional Bladder BMD = 0.524 (1.44) Liver Thyro id 1000 Lung Rat Mouse 100 10 Tumor BMD Tumor BMD (mg/kg/d) 1 r = 0.940 (p = 0.0002) 0.1 0.1 1 10 100 1000 10000 Lowest Transcriptional BMD (mg/kg/d) Thomas et al., Toxicol Sci, 2013 National Center for Computational Toxicology Why Could this Be True? • Most histological changes do not occur without upstream or downsream changes in the transcriptome • Most environmental chemicals are highly non-selective in their interactions with biological systems Thomas et al., Tox Sci., 2013 National Center for Computational Toxicology Integration of TGX in a Tiered 21st Century Toxicity Testing Framework Protective Predictive Protective Predictive 15 National Center for Computational Toxicology The Next Frontier for Toxicogenomics ToxCast ToxCast ~600 assays ~1,000 chemicals Tox21 Tox21 ~50 assays ~10,000 chemicals Response 16 Concentration National Center for Computational Toxicology Developing a Portfolio of High- Throughput Toxicogenomic Tools High-Throughput Transcriptomic Screening Platform (TempOSeq) • Low cost, 384-well, cell lysate compatible • Whole transcriptome (EPA), S1500+ (NTP) • Workflow integration of reference materials and controls, development of performance standards • Portable platform/workflow for collaborative data generation 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 A MAQC-A (Us) 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 non-treated B MAQC-A (Us) 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 non-treated Platform SelectionC MAQC-B (Us) 10 10 10 10 10 10 10 10Workflow10 10 10 10 Verification10 10 10 10 10 Study10 10 10 10 10 non-treated D MAQC-B (Us) 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3