Predictive modeling, that is, the process of developing a mathematical tool or model that generates an accurate prediction, is the topic in a new book just published by Springer titled Applied Predictive Modeling, by Kuhn and Johnson. Pre-clinical drug discovery and development is focused on predicting the extent to which a chemical will be safe and efficacious once dosed in man. Many of the predictive methods normally used involve measurements of responses in biological assays such as cell cultures or animal models. As more and more experimental data are generated in this manner, it may in some cases be possible to capture the relation between stimuli (e.g. dosing of a chemical), and the experimentally measured responses, in a mathematical model (we recently developed a predictive model for oligonucleotide-induced hepatotoxicity in mice, which you can read about here if interested). The book by Kuhn and Johnson therefore seems to be an interesting read for mathematically-inclined drug hunters. While predictive modelling naturally cover hallmark bioinformatics disciplines such as machine learning, pattern recognition, and data mining, which have been treated in many books already, in this book, the entire process is
the focus, which seems interesting. Naturally, the statistical programming language chosen for the examples is R. Happy Christmas vacation reading – we’ll hopefully update this blog entry in the new year with our reading impressions (there is also a nice review of the book on R-bloggers).