Project 4: Methods for subtype detection in high-dimensional data with a special focus on longitudinal data

Project description

With biological data sets becoming bigger, the interest in applying machine learning to precision medicine, to improve diagnosis/prognosis and personalise treatment has increased substantially. Like many others, psychiatric disorders are heterogeneous and its definition is still extensively based solely on symptoms.

Consequentially, we aim to explore and develop methods for unsupervised learning on genetics and high-dimensional longitudinal data coming from different sources, such as DNA methylation, fMRI studies and animal motion tracking, which has the potential of leading us to better data-driven disease characterization.

To make our results applicable to a research audience as wide as possible, we are also interested in deploying our methods to the community as software packages. For this, we focus on usability, scalability and speed of the implemented algorithms as well.
Given the nature of our institution the original plan sticks to psychiatry-related data, the methods and algorithms to be developed are immediately applicable in other disease areas and data domains.

Secondments

1. Clustering patients for improved patient state modelling network structures
   Host: ETH Zürich
2. Learning patient-subgroup specific representations from big data
   Host: Pharmatics
   Planned date: December 2021 – February 2022

Activities

• EUvsVirus Hackathon

Publications

1. Lucas Miranda, Riya Paul , Benno Pütz, Nikolaos Koutsouleris and Bertram Müller-Myhsok. (2021). Systematic Review of Functional MRI Applications for Psychiatric Disease Subtyping. Frontiers in Psychiatry 2021, 12:665536. doi: 10.3389/fpsyt.2021.665536.
2. Lucas Miranda, Joeri Bordes, Serena Gasperoni, and Juan Pablo Lopez. Increasing resolution in stress neurobiology: from single cells to complex group behaviors. Stress, 26(1). https://doi.org/10.1080/10253890.2023.2186141