Prediction models are a cornerstone of clinical care as they enable (early) diagnosis of disease or personalized treatment.
Machine learning (ML) methods hold the promise to improve upon classical models, but for these to be useful in clinical practice they need to be tailored to (small
n) medical settings. Moreover, ML prediction models need to be complemented with interpretability tools, and allow uncertainty quantification. This project contributes to these aims by applying learners that can incorporate external information, and by developing dedicated interpretability tools that allow to assess and infer the importance of (groups of) features and their interactions. As many of the applications have a genomics component, an extra challenge will be to deal with high-dimensional (
p > n) data. The project entails methodological development, balanced with implementation and applications in the onco- and neurological field. You will have access to high-quality clinical data from our collaborators.
Would you like to know more about the different phases within the PhD trajectory? You can read more about this on this page.Your main role is to develop and implement methods to quantify and test feature importance for machine learned predictions. You will use available tools for including external information to optimize performance of these learners. In addition, you will critically apply and test these methods to onco- and neurological data sets in collaboration with medical researchers. Results will be published in both methodological and biomedical journals. Finally, you are expected to play an active role in the
ADORE research community, in particular the
Biocomputational Focus group.