S3: How to Build a Predictive In Vitro Method?
Institute of Bioengineering and Nanotechnology (IBN)
Principal Research Scientist
Predictive in vitro methods are crucial for 21st Century Toxicology. Currently, most in vitro methods are not predictive or their predictivity has not been determined. With respect to predictive assays there is often substantial confusion about the question what an assay actually predicts, and the way how results are presented is often not helpful. Crucial for developing predictive methods is pre-validation with a statistically robust set of compounds, and proper compound annotation and classification require careful consideration, which is still difficult to achieve with data mining procedures. Whereas most discussions focus on selection of proper cell and tissue models, little attention is paid to endpoint selection. Examples will be discussed how bioinformatics methods can be used for the identification of proper endpoints. In addition, the analysis of high-throughput data will be addressed.