If you are confronted with incomplete data for a regulatory submission, the data gaps can be filled with predictive modeling. The U.S. E.P.A., Environment Canada, and the European Chemical Agency will accept predicted data, provided some conditions are met.
Some modeling tools are well-established and generally accepted by regulatory agencies and require little additional justification - for example, ECOSAR is used to predict logKow by most regulatory agencies.
On the other hand, there are no established models for predicting many physical chemistry properties or toxicological endpoints. However, using tools such as the OECD QSAR Toolbox, a model can often be built based on available data for similar chemicals within that class. In order for the predicted value to be accepted, the model needs to fulfill several criteria:
it has to predict a well-defined endpoint
the algorithm has to be unambiguous
the applicability of the domain has to be justified
there has to be appropriate measures of data robustness and predictive power
if possible, a mechanistic interpretation
CERM has extensive experience with all the predictive tools used by EPA - ECOSAR, OncoLogic, and ChemSTEER - and have both trained people to use them and helped clients trouble-shoot situations where the models may not be straight-forward. CERM also has experience using the OECD QSAR Toolbox to help clients fill data gaps and avoid expensive testing.
For more information please contact Peter Ranslow.