General Hospital of Vienna
Venue: Lecture Hall, Room 21, Level 8
Time: Tuesday, March 18th, 14:30 – 16:00
Hosts: Michael Trauner, Thomas Scherer, Cecile Philippe, Alexandra Kautzky-Willer
“GenAI that can reason about biomedical data”
Topic
Understanding biological systems at single-cell resolution requires computational models that disentangle true biological signals from technical artifacts and integrate multimodal data. In this talk, I will discuss three recent advances: EmptyDropsMultiome, which improves nucleus detection in singlecell multiomics; Celcomen, a causally grounded generative model for counterfactual predictions in spatial transcriptomics; and CellDISECT, which leverages variational autoencoders to model multibatch, multi-covariate single-cell data. Finally, I will outline future work on a Vision-Genomics Foundation Model integrating histology and spatial transcriptomics to study emergent biological phenomena, with applications in aging, tissue architecture, and disease modeling.
Biography
Stathis studied Physics at the University of Athens and CERN where he created and
analysed simulations of supersymmetric field theories. He then completed his MSc and PhD at UCLA on physical mathematics and theoretical physics, focusing on defining rigorous and exactly solvable models of 2d topological quantum gravity equipped with any open/closed Atiyah TQFT and using them to derive insights about the black hole information paradox. Stathis joined the Teichmann lab in March 2023 as a postdoc, working closely with the Lotfollahi lab. He is interested in the mathematical design of deep learning algorithms for disentangled representations and causal inference in single-cell and spatial genomics, with an aim at understanding gene regulation and protein-ligand interactions.
