We are happy to announce our third online seminar in the Biostatistics Seminar Series on Thursday, November 24th, 16h-17h.

This series of Biostatistics seminars targets a broad (bio)statistical audience, in particular PhD-students. Specialists discuss a topic of their interest, paying particular attention to concepts relevant and accessible to a non-specialist audience as well.

Speaker
Maren Hackenberg
Institute of Medical Biometry and Statistics, Faculty of Medicine and Medical Center
University of Freiburg, Germany

Title
“From dynamic modelling of disease trajectories to single-cell omics: state-of-the-art methods development in Julia”

Abstract
Over the years, many software tools have been developed for statistical modelling and scientific computing. While they often share a set of similar characteristics, each approach typically carries a legacy to the context in which it has originally been developed. For example, systems like R and SAS have their origins as convenient interfaces to numerical libraries and have therefore not been optimised for implementing complex algorithms in the language itself, while solutions such as Python come from general-purpose computing and have only been extended post hoc for statistical modelling. As a rather young language, Julia is not burdened by such legacies. This makes it particularly suitable to pick up new emerging modelling approaches such as differentiable programming, which, e.g., require straightforward access to automatic differentiation frameworks.

In my talk, I will introduce Julia as a fresh and pragmatic approach to scientific computing and show in several examples how the low-barrier approach of Julia allows to more readily implement scientific ideas that researchers might not have pursued with other tools due to runtime inefficiencies of missing flexibility of libraries. Specifically, I will focus on differentiable programming for flexibly integrating different modelling paradigms, such as neural networks and differential equations. I will show how combining deep learning with dynamic modelling allows for describing individual-specific disease trajectories in a rare disease registry of patients with spinal muscular atrophy, where longitudinal data on patients’ disease developments is collected during routine visits and only a relatively small number of patients and few and irregular follow-up time points are available. Further, I will illustrate how Julia facilitates interpretable modelling of single-cell RNA-sequencing data using deep learning.

With this, I will argue that Julia removes barriers that so far limit scientific ideas in the modelling community and that by adding Julia to your toolbox, you can expand the range of modelling approaches available to your research.

Note: Registration is not required.