My work translates biological and clinical questions into quantitative form. Mechanistic models where the biology is understood; statistical and machine learning models where it is not; and uncertainty quantification to keep the boundary between the two honest.
Systems medicine at Universitätsklinikum Carl Gustav Carus Dresden and TU Dresden.
PräMo · Uniklinikum Dresden
Part of Präzisionsmedizin und Molekulare Prävention (PräMo). The question: can changes in immune cells and blood proteins reveal the early signs of metabolic disease before a clinical diagnosis? I analyse cross-sectional and longitudinal immunometabolomic data to trace how immune and metabolic signals interact along the trajectory towards type 2 diabetes, working with the UK Biobank alongside in-house cohorts.
IRTG 3019: MEDIS
Does infection risk begin to rise before blood glucose reaches the clinical diabetes range? This project examines whether risk is already elevated at earlier stages of glycaemic dysregulation, and uses plasma proteomics to identify proteins associated with infection outcomes — pointing at the immune pathways that might connect glycaemic status to infection susceptibility.
Deliberately mixed — no single method answers a clinical question on its own.
Proteomic, metabolomic and clinical data integrated across layers, cross-sectionally and longitudinally, in large biobank-scale cohorts.
Supervised models for time-to-event and risk prediction, Bayesian inference for parameter estimation, and calibration against sparse, noisy clinical data.
Reaction–diffusion systems, agent-based and hybrid discrete–continuum models of tumour growth, invasion and immune interaction.
Quantitative feature extraction from MRI and microscopy, segmentation pipelines, and feature selection that survives small clinical cohorts.
Global sensitivity analysis and uncertainty quantification — which parameters actually drive a prediction, and how far it can be trusted.
FAIR, ethically governed analysis pipelines built to be re-run and audited, plus high-performance computing for large parameter sweeps.
TU Dresden, 2021–2025 — supervised by Prof. Haralampos Hatzikirou.
Biopsy is standard of care, but used mostly for diagnosis rather than prediction. Using a spatio-temporal model of tumour–immune interaction I generated a cohort of virtual patients and “observed” clinically realistic outputs — MRI volumes, Ki-67 proliferation, biopsies at different locations. Macrophage density at the tumour edge turned out to carry substantial predictive value for time to relapse.
My thesis, Modeling Cell Plasticity at the Invasive Edge to Control Glioblastoma Recurrence, developed hybrid in-silico frameworks for glioma–macrophage interaction, ran them as in-silico clinical trials over virtual cohorts, and combined mechanistic modelling with machine learning to attack the resulting inverse problems.
Computational modelling of how chemotherapeutics distribute through heterogeneous, vascularised tumour tissue, including spatio-temporal studies of doxorubicin.
In-silico studies of magnetic nanoparticle transport in tumours with necrotic regions — how particle size, field intensity and dynamic microvasculature govern where the dose lands.
CFD versus fluid–structure interaction predictions in a patient-specific giant saccular cerebral aneurysm, with the Biological Engineering Lab at the University of Tehran.