Sarah Filippi

I have a broad interest in methods in statistical modelling and machine learning, their theoretical properties and their application to biomedical problems. I am particularly interested in Bayesian statistics and nonparametric methods.

The focus of my research group lies in developing statistical machine learning and computational statistics methods. Some of these methods are motivated by applications in and around computational biology, biomedical genetics, pharmacology, epidemiolgy and clinical studies. We are particularly interested in addressing how novel statistical and computational approaches can aid in the analysis of large-scale real-world health problems and in the understanding of complex biological systems in health and disease.

In terms of methodological development, we focus on a breadth of approaches including:

• Bayesian non-parametric approaches for causal inference, pattern recognition as well as variable importance and model selection in high-dimensional settings
• Decision making process under uncertainty (from Reinforcement Learning to experimental design and Bayesian optimization)
• Kernel methods and scalable Gaussian Processes

In collaboration with experts in the application domains (such as biologists, epidemiologists, or clinicians), we adapt and tailor state-of-the-art methods to various types of real application settings in biomedicine including:

• Variable importance to identify genes related to survival; relevant OTUs in microbiome
• Model development and Bayesian inference of complex longitudinal models (such as the evolution of chronic diseases, or mechanistic models of cellular dynamics in health and their perturbation in disease)
• Incorporating uncertainty quantification in chemometrics models and their impact in decision making in pharmacology
• Better understanding of the underlying mechanisms of some chronic diseases such as cancers, diabetes, cardiovascular diseases, or respiratory diseases (e.g., asthma) and identifying causal (genetic or environmental) risk factors or diagnostic tools