Can public data improve mutagenicity predictions for proprietary compounds.pdf
It has long been axiomatic that predictive toxicology models built on data available in the public domain can encounter problems when attempting to predict proprietary data, e.g. from a drug development program. Conversely, model performance against public data is often used to support the selection of models for specific applications and such analyses are often published. One outcome of some recent model development work has been to produce metrics that can be used to experimentally test the axiom and assess the validity of basing conclusions on public data.
Presented by Chris Barber at SOT, Phoenix, USA; 24th March 2014.