Quality requirements for activity and effects data

A comprehensive set of substance-specific data for environmental hazard assessment requires information on well-defined endpoints concerning 

The data-quality requirements for QSAR models relate to several aspects of the experimental procedure, data transformation and the selection of the appropriate test compounds. Only if the input data of a QSAR meet the highest quality standards may a sound model be derived. Because the accuracy of predictions can never be better than the variability of the respective measurements (usually 20% and more), validity assessment of the activity and effects data is crucial in QSAR derivations. The data should be generated by tests that are methodologically and mechanistically defined. The latter is not trivial for parameters such as biodegradability, soil sorption and ecotoxicity. With regard to the considerable variability of measurements, inter- and also intra-laboratory, the test results, especially when collected from different literature sources, should be critically evaluated with respect to  

The experimental data underlying any QSAR model should preferably be determined by standardized protocols, but the absolute values will still vary in reliability and precision, which must affect the accuracy of the predictions based on the acquired QSARs. Besides sufficient precision and reliability, the activity and effects data for QSAR analyses should principally meet these additional requirements  

The compounds subject to a QSAR analysis must differ sufficiently in activity to provide the basis for a representative model and to avoid the description of local phenomena. If the activity range is too narrow it may over-value the inherent variance in the experimental data, which may significantly exceed 20% for parameters such as soil sorption, bioconcentration and ecotoxicity. The difference in activity between compounds must therefore clearly exceed that among the replicated measurements for each of the test substances. Otherwise, the derived QSAR may show an artificial relationship between the compounds structure and the experimental scatter, but not the effects under investigation. 

Identical mode of action must yield identical QSARs. 

---