May 4, 2016
Using In Silico Software to Generate an ICH M7 Submission
and Expert Review
GTA WORKSHOP
Workshop Chairs
• Chris Barber • Catrin Hasselgren • Roustem Saiakhov
Facilitators
• Suman Chakravarti• Kate Kearney• Richard Williams• Laura Wirpza• Glenn Myatt • Kevin Cross
AGENDA
• The Basics• Introduction to Hands on Exercises• Work on Case Studies• Perspectives from MultiCASE, Lhasa and
Leadscope• Wrap Up
Expert Rule Based vs. Statistical Methods
• ICH M7 requires use of 2 complimentary systems to maximize coverage
• Statistical – built by statistical mining of training data sets– Build “from scratch” or modify models with new data– Expert input - choice of descriptors and algorithm
• Expert Rule Based– Expert relates structural features to toxicity– Data curated from literature or generated to support rule
development– Supported by examples, a proposed MOA, explanation of
the scope of the rule– Statistical mining of data – define mitigating factors
Choosing Systems
• System that helps you make an expert conclusion– Coverage of your chemical space– Supporting information to aid expert review of
predictions• Compliant with OECD principles
– e.g defined domain of applicability• Accepted by regulators
– Well characterized, publications, used by or familiar to regulators
Choosing Systems
• “Makes my life easier” considerations– Presentation of supporting information– Availability of supplemental information– Easy access to literature references, supporting
databases– Batch loading – Generation of reports for internal use – Generation of reports for regulatory submission– Good technical support
Combining Results – Overall Prediction
Combining Results – Overall Prediction
Expert Judgment
When Is Expert Review Necessary?
• ICH M7 says expert review “if warranted”• Experience says – Always! It adds value• Amount of effort will vary
2 Systems Negative
ConflictingEquivocal
OOD
Expert Review – The Value
• Any positive prediction concluded mutagenic– Simple and conservative….BUT– Drives control or Ames testing– May lead to unacceptable # of false positives
• 2 Negative predictions concluded not mutagenic– Expert review improves negative predictivity– Identify reactive features “missed” by 2 systems
• Out of domain predictions– Identify compounds that can support your conclusion
• Expert review of published data
“How To” Conduct Expert Review
• Details, supported by practical examples– Barber et al., Reg Tox Pharm 2015. 73(1), 367– Powley. Reg Tox Pharm 2015. 71(2),295 – Amberg, et al., Reg Tox Pharm 2016. 77, 13
• Some systems guide user through expert review process
• Critical review of information supporting each prediction
“How To” Conduct Expert Review
• Negative predictions– confirm query chemical is within applicability domain– No structural features that suggest reactivity
• Check validity of mitigating factors for alerts
• Positive predictions– Confirm relevance to query chemical
• Statistical – check supporting training set chemicals for alert– Run training set chemical through expert systems to identify
other more likely causes of activity
• Alert’s structural environment – is it relevant?• Similarity to API or other chemicals reported not mutagenic
Regulatory Submission – Expert Review
• What and how much information should be provided re: expert review?– Convince yourself → convince others
Regulatory Submission – Expert Review
• Conclusion consistent with predictions– Detailed description of expert review not needed
• Conclusion is different from the prediction(s)– Strong scientific rationale
• Steric influences, reactivity, more knowledge of SAR• Experimental data on other related compounds• Supporting literature references
Regulatory Submission Information
• (Q)SAR systems used and version numbers• For each structure
– Experimental data if it exists– (Q)SAR results for each system– Overall conclusion– Class 1 – 5 assignment– Information relating to expert review
• Market application– Bacterial mutagenicity study reports are required– Reports from systems not required but may be
included.
Next Up….HANDS ON EXERCISES