Data analysis methods for systems medicine

Our group focus on developing novel methodologies for systems medicine problems.  Currently, we are working on the following methodologies:

 

“Prognostification” of medical images 

In clinical practice, imaging and biopsy sampling are the most important diagnostic tools. Analysing these “snapshots” in time, doctors attempt to predict the future of the lesion and propose an appropriate treatment. However, there is no way to understand a single scene photograph of a Shakespeare play without knowing the story behind it. Likewise, microscopic biological “scenes”, of a cellular ensemble expressing various phenotypes, can have true prognostic power if interpreted in the context of the underlying dynamic processes of cell phenotypic decisions and interaction with microenvironmental/stromal entities. Our method has been applied to unravel the prognostic potential of immune infiltration patterns as found on breast tissue biopsies.


Grey-box methodology for integrating –omics in agent-based models

Agent-based spatiotemporal models (ABM) have been proven very useful in simulating tissue dynamics. Each cell is described as a single agent and a balance of forces determines its position. Typically, each cell includes a stochastic decision-making mechanism that model’s phenotypic changes according to environmental cues. However, their integration with –omics data has not been successful - apart from metabolomics. This occurs due to the lack of knowledge concerning the cell’s mechanisms that translate -omics to phenotypes. Currently, we combine machine learning and dynamic modeling to develop a comprehensive multiscale framework.