Good Read-Across Practice 1: State of the Art of Read-Across for Toxicity Prediction Mark Cronin Liverpool John Moores University England
Acknowledgement
What I am Going to Say Background and context State of the art of read-across Practical issues Quantification Supporting read-across Tools Guidance Case Studies Acceptance Good Read- Across Practice These are my views and others may wish to dissociate themselves from them
Category Formation (Grouping) for Read Across Read-across uses information from members of a group of similar compounds, with known activity, to predict activity of unknown(s) OH OH OH OH Toxicity Toxicity SAR / Read- Across Interpolation
Some Good Reasons for Using Read-Across Its simple, cheap and transparent It has regulatory acceptance (if done correctly) Provides solutions to problems
Potential Uses of Read-Across REACH and other global legislation New and existing chemicals Prioritisation, C&L, Risk Assessment New product risk assessment (e.g. industry) New product registration AOP / IATA Framework Nanomaterials Pharmaceuticals development Pharmaceuticals impurities Legal highs / illicit drugs Others
Example of a Category: Long-Chain Alcohols Veenstra G et al (2009) Ecotox Environ Saf 72: 1016-1030.
Example of a Category: Terephthalic Acid and Esters Ball GL et al (2012) Crit Rev Toxicol 42: 28-67.
QUESTION: Can We Fill These Data Gaps? Ball GL et al (2012) Crit Rev Toxicol 42: 28-67.
Can We Fill These Data Gaps? Probably. If we have. High quality source data Consistency within the data for the category We are interpolating There is a good reason and justification for data gap filling We can demonstrate similarity Ball GL et al (2012) Crit Rev Toxicol 42: 28-67.
Well Known, and Worrying, Activity Cliffs Exist Which Demonstrate Problem of Identifying Similar Compounds Teratogen Sedative
Are These Similar Molecules? Fingerprint Tanimoto maccs 0.77 fp4 0.67 fp2 0.64 fp3 0.50
Similarity is More Than Similarity in Chemical Structure Non Sensitiser Strong Skin Sensitiser
Guide to Grouping Chemicals Structural Analogues OH OH OH OH O Mechanistic Analogues O OH N O Mode of Action Analogues O OH OH HO HO HO
Other Options for Grouping Chemicals Compounds that are metabolised to a common molecule Compounds that are degraded rapidly to common products Metrics of molecular similarity
Like it or Not The Use of Read-Across is a Reality Toxicity to reproduction (1 882 dossiers covering phase in substances 100-1 000 tpa) Developmental toxicity (1 882 dossiers) Full Report Published 2 June 2014 available at: http://echa.europa.eu/documents/10162/13639/alternatives_test_animals_2014_en.pdf
Growth in Publications Web of Science literature search using key words Read- Across and Toxic performed 20 February 2016
Like it or Not The Use of Read-Across is a Reality However the acceptance of readacross predictions is not fully known Toxicity to reproduction (1 882 dossiers covering phase in substances 100-1 000 tpa) Developmental toxicity (1 882 dossiers) Full Report Published 2 June 2014 available at: http://echa.europa.eu/documents/10162/13639/alternatives_test_animals_2014_en.pdf
Key Issues with Read-Across How can we support a read-across prediction? i.e. provide further (biological) evidence that chemicals belong to a group When do read-across predictions become acceptable to replace an animal test?
State of the Art of Read-Across and Good Read-Across Practice
Good Read-Across Practice Practical Issues with Undertaking Read-Across Similarity is a Simple and Fundamental Concept Difficult and Subjective Confirmation and Evidence Required Assuring Category Membership Well recognised approaches Much guidance Consideration of endpoint to identify best approach Proof is essential for regulatory acceptance Identification and reduction of uncertainties Support from New Methods Data / Biological Similarity
Practical Issues with Undertaking Read-Across Good Read-Across Practice Can be Described R-A Arguments, Data, TK etc Defining Uncertainty Some uncertainty, context dependent, must be considered acceptable Criteria for defining uncertainty proposed but not necessarily accepted Further effort required
Practical Issues with Undertaking Read-Across Good Read-Across Practice Various Resources Assessing and Assigning Quality Biological Data Increasing data availability e.g. echemportal, ECHA DB etc New methods data e.g. HTS, Tox21 Biological profiling will support read-across
Good Read-Across Practice Specific Use Case Scenarios Many Case Studies Difficult to Confirm No Toxicity Confirming the Presence of Toxicity No / Low Toxicity Good examples for e.g. reactive toxicity Some areas more effort e.g. receptor mediated toxicity Few robust categories, map onto OECD / HPVC? Similarity in toxicokinetics may need to be proven Effort needed in using biological similarity Other Area: Nanomaterials, Mixtures, UVCBs
Supporting Mechanistically-Based Read-Across Good Read-Across Practice Clear Linkages to Category Formation Supports Hypothesis of Toxicity AOPs Molecular Initiating Events form the basis of grouping Data from assays for key events may confirm category membership Data from key events may be quantitative May form the basis of ITS / IATA, case studies required
Good Read-Across Practice Quantification of Read-Across Qualitative R-A is the current norm Some examples Very important, little addressed Few data How to Quantify R-A Toxicokinetics Appreciation of (PB)PK modelling will be required Effort needed on how to incorporate new methods data More understanding, e.g. through case studies, is needed Requires more data and understanding May support quantification, similarity assessment
Chemoinformatics: Tools for Grouping, Databases, Predictions of Toxicity, Metabolism etc
Tools and Databases Not An Exhaustive List Tool Grouping Tox Data ADME Mechanism Free Tox/Track Yes Yes Partial Some Yes Many bespoke DrugMatrix tools for grouping and read-across May need further guidance / illustrated case studies Yes Few Some Yes Few No Yes Yes
Tools and Databases Not An Exhaustive List Tool Grouping Tox Data ADME Mechanism Free Tox/Track Yes Yes Partial Some Yes DrugMatrix Yes Few Some Many data sources support read-across Always opportunities for further data sharing Yes Few No Yes Yes
Tools and Databases Not An Exhaustive List Tool Grouping Tox Data ADME Mechanism Free Tox/Track Yes Yes Partial Some Yes DrugMatrix Yes Few Some (Quantitative) metabolite and PK property prediction requires development and better integration into read-across Yes Few No Yes Yes
Tools and Databases Not An Exhaustive List Tool Grouping Tox Data ADME Mechanism Free Tox/Track Yes Yes Partial Some Yes DrugMatrix Yes Few Some A mechanistic basis to read-across is desirable AOPs may support read-across in a number of ways Yes Few No Yes Yes
Current Guidance Many sources Need for consistent approach to reporting and assessing read-across Adoption of ECHA s Read-Across Assessment Framework (RAAF) and ensure effectiveness Several Other Initiatives
Four repeat dose RA case studies Case Studies Many examples Need for more to address issues such as RAAF, uncertainty, reporting, biological profiling etc Ten safety assessments using RA Several Other Initiatives
Acceptance of Read-Across Variable Addressed in next talk (Some) Key Points Getting the documentation right Read-across argument Acceptable level of uncertainty
Conclusions Practical issues affecting read-across have been identified, if not resolved Useful tools and databases Much guidance and opinion Less certainty about certainty Acceptance variable
Acknowledgements The European Community s Seventh Framework Program (FP7/2007-2013) COSMOS Project under grant agreement n 266835 and Cosmetics Europe The CAAT GRAP Drafting Groups Co-workers in Liverpool, EU, USA