Incorporating new technologies and High Throughput Screening in the design and selection of chemical alternatives
Download as PPTX, PDF0 likes304 views
The document outlines advances in chemical design and toxicology through high-throughput technologies and computational models, aiming to reduce reliance on animal studies. The U.S. EPA's National Center for Computational Toxicology is central to integrating data and technologies for better predictive models across chemicals. Progress is made through extensive datasets and screening methods to facilitate predictions for safety and bioactivity of thousands of chemicals.
Incorporating new technologies and High Throughput Screening in the design and selection of chemical alternatives
- 1. Incorporating new technologies and high- throughput approaches in the design and selection of chemical alternatives Antony Williams U.S. Environmental Protection Agency, RTP, NC June 15th, 2017 21st Annual Green Chemistry & Engineering Conference, Reston, VA This work was reviewed by the U.S. EPA and approved for presentation but does not necessarily reflect official Agency policy. http://www.orcid.org/0000-0002-2668-4821
- 2. Paradigm Shift in Chemical Design (especially in regards to toxicology) • Chemical Design in the future – High-throughput modeling – SAR/QSAR/QSTR models – Retrosynthetic analysis algorithms for synthesis • Toxicology – FEWER animal studies. Maybe NONE?? – Openness in data sharing – More confidence in computational predictive models that link chemical structures to adverse outcomes 1
- 3. It will depend on lots of DATA 2
- 4. What we need to progress • Adequate information on potential bioactivity for as many chemicals as possible • Read-across, QSAR and QSTR models from large high-quality curated datasets 3
- 5. • National Center for Computational Toxicology established in 2005 to integrate: – High-throughput and high-content technologies – Modern molecular biology – Data mining and statistical modeling – Computational biology and chemistry • Currently staffed by ~60 employees as part of EPA’s Office of Research and Development • Home of ToxCast & ExpoCast research efforts • Key partner in U.S. Tox21 federal consortium • Multiple cross-division collaborations National Center for Computational Toxicology
- 6. What we need to progress • Adequate information on potential bioactivity for as many chemicals as possible • High Throughput Screening ToxCast/Tox21 generating and making available bioactivity data • GREAT progress to date – more to come! • Read-across, QSAR and QSTR models from large high-quality curated datasets • Chemistry data gathering and high-throughput bioassays are providing the necessary data to generate these models 5
- 7. Data and Modeling Efforts to Date ToxCast ExpoCast Predictive Models Comp Chemistry HTTK • High-throughput in vitro screening of ~4,000 chemicals across ~1200 assay endpoints (ToxCast) and ~8,000 chemicals in ~60 assay endpoints (Tox21) • High quality, curated chemical structure and physical chemical properties database of ~750,000 chemicals • High-throughput exposure estimates with uncertainty for ~ 7,000 chemicals based on production volume and chemical use
- 8. Data and Modeling Efforts to Date • High-throughput in vitro screening of ~4,000 chemicals across ~1200 assay endpoints (ToxCast) and ~8,000 chemicals in ~60 assay endpoints (Tox21) • High quality, curated chemical structure and physical chemical properties database of ~750,000 chemicals • High-throughput exposure estimates with uncertainty for ~ 7,000 chemicals based on production volume and chemical use • Legacy in vivo data from ~6,000 animal toxicology studies on ~1,110 unique chemicals • High-throughput toxicokinetic (HTTK) models for ~700 chemicals based on in vitro measurements • AOPs and virtual tissue models for broad range of developmental toxicities ToxCast ExpoCast Comp Chemistry HTTK ToxRef Predictive Models
- 10. Approximately 15 Years of Data… Growing with daily curation DSSTox_v2 PublicCurated 5. Low 2. Low 6. Untrusted 1. High 3. High 4. Med 7. Incomplete Public_Untrusted Public_Low Public_Medium Public_High DSSTox_Low DSSTox_High 4535 16K 33K 101K 584K ~ 310K pending ~ 150K pending
- 11. Features Properties Chemotypes, fingerprints, physchem properties, ... Chemical representation levels supporting data integration 10 Structure Generic Substance Test Sample Chemical Name CASRN Supplier, Lot/Batch, physical description SMILES InChI
- 12. Features Properties Chemotypes, fingerprints, physchem properties, ... Chemical analogs, Read-across, SAR modeling Chemical representation levels supporting data integration 11 Structure Generic Substance Test Sample Chemical Name CASRN Supplier, Lot/Batch, physical description SMILES InChI Experimental Endpoint Data Public toxicity datasets Structure searching & modeling
- 13. Integrating in vitro and in vivo data HTS-In vitro ToxRef COSMOS ToxCast Tox21 Chemicals DSSTox Measured Properties Chemical KB Biological Data Biological activities
- 14. Adding Product Use and Exposure HTS-In vitro ToxRef COSMOS ToxCast Tox21 Chemicals DSSTox Measured Properties Product Use & Exposure Chemical KB Biological Data Biological activities
- 15. High Throughput Measurement to Characterize Exposure 14
- 16. Building Models from the data HTS-In vitro ToxRef COSMOS ToxCast Tox21 Chemicals DSSTox Biological Data Fate & Transport ADME Reactivity Biotransformation Physchem properties Biological activities Measured Properties ModelsExposure Data Toxicity Predictions Structure-alerts Chemical KB Product Use & Exposure
- 17. Access to Our Data and Models 16
- 18. What we have learned… • Data curation, standardization and versioning is essential • Prototype application development suffices for research projects but production development requires managed processes • ODOSOS (Open Data, Open Source and Open Standards) endows many benefits • With this in mind… 17
- 20. Bisphenol A 19
- 22. Chemical Properties 21 • Consuming and producing open data
- 23. Data Distribution 22 • Consuming and producing open data
- 25. Modeling Details 24 Calculation Result for a chemical Model Performance with full QMRF Nearest Neighbors from Training Set 24
- 26. Prediction Details and QMRF Report 25
- 27. Developing “NCCT Models” • Our approach to modeling: – Obtain high quality training sets – Apply appropriate modeling approaches – Validate performance of models – Define the applicability domain and model limitations – Use models to predict properties across our full datasets • Release as Open Data and Open Models
- 28. Workflow Details and Data 27 OPERA Models: https://github.com/kmansouri/OPERA
- 29. Making Toxicity Value Data available 28
- 30. Exposure 29
- 31. Consumer Product data 30 • Data gathered from multiple online data sources – text mining and data downloads • CASRN and name mappings to produce curated structure set with functional uses • Produce Functional Use (FuseDB) and build predicted functional use models
- 32. NHANES Exposure data 31 Environ. Sci. Technol., 2013, 47 (15), pp 8479–8488
- 33. ToxCast and Tox21 Bioassays 32
- 34. ToxCast and Tox21 Bioassays 33
- 35. PubChem Bioassay Data Integration 34
- 36. PubMed: BIG DATA Literature Search Integration 35
- 37. Linked Directly to PubMed 36
- 38. Literature Search Integration Google Scholar 37
- 39. Batch Data Searching • I have a 5000 CAS Numbers (or Names) – is there data available? – Has any Toxcast data been run? – Are there Toxicity Data values available? – Are there predicted exposure data? – Can I get predicted physchem data for my model? 38
- 40. We get lots of user feedback.. • Online since April 2016 • Feedback welcomed! 39
- 41. “RapidTox” in Development • Semi-automated decision support tool for high-throughput risk assessments • Use Dashboard “architecture”, existing data streams and add new data – e.g. Global Hazard Summary and ECHA data • Combine data streams into quantitative toxicity values with uncertainty estimates
- 42. Risk Assessments Generally Contain a Standard Set of Components We are assembling these components to deliver RapidTox Phys Chem Exposure Hazard Dose Response, PK, and PODs Risk Summary Uncertainty Variability
- 43. Quantifying Uncertainty & Assessing Performance of Read-Across - GenRA • “Chemical-Biological Read-Across” - predict toxicity as a similarity-weighted activity of nearest neighbors • Evaluates read-across performance and uncertainty using available data
- 45. We’re not done yet… HTS-In vitro ToxRef COSMOS ToxCast Tox21 Chemicals DSSTox Biological Data Fate & Transport ADME Reactivity Biotransformation Phys-chem properties Biological activities Measured Properties ModelsExposure Data Toxicity Predictions Structure-alerts Chemical KB Product Use & Exposure
- 46. Acknowledgements EPA-RTP Russell Thomas Kevin Crofton Chris Grulke Ann Richard Richard Judson John Cowden John Wambaugh Imran Shah Grace Patlewicz and… Many other contributors from across EPA-ORD
- 47. Contact Antony Williams US EPA Office of Research and Development National Center for Computational Toxicology (NCCT) [email protected] ORCID: https://orcid.org/0000-0002-2668-4821 46