The power of combining
09 Oct 2012
Being able to predict the clinical outcome in individual breast cancer patients is essential to ensure each patient gets the right therapy (and is spared unnecessary treatment). Both clinical data and gene expression data are used to derive predictors of clinical outcome. However, combining the two sources to build a single prediction model remains rare and the few examples where it has been tried have yielded inclusive results on prediction performance. Together with colleagues from the Netherlands Cancer Institute and the Academic Medical Centre Amsterdam, Martin van Vliet (Delft University of Technology) performed an extensive comparison of three integration strategies as well as using five classifiers of varying complexity without integration data types. They used a dataset of breast cancer samples of which both gene expression profiles and clinical data are available to develop predictors, which were subsequently validated using four independent breast cancer datasets. Their main conclusions are that combining the two data types increases the performance of all five classifiers and that the late OR integration strategy generates the best overall result of the three strategies tested.
van Vliet MH, Horlings HM, van de Vijver MJ, Reinders MJ, Wessels LF
Integration of clinical and gene expression data has a synergetic effect on predicting breast cancer outcome.
PLoS ONE 7(7): e40358. doi:10.1371/journal.pone.0040358<br />