Multi Scale Generative Modeling of biology: prior knowledge informed Neural Cellular Automata of tissue self-organization
Event details
| Date | 04.12.2024 |
| Hour | 14:15 › 15:30 |
| Speaker | Varun Sharma |
| Location | |
| Category | Conferences - Seminars |
| Event Language | English |
Biological systems exhibit emergent properties that arise from complex interactions across multiple organizational scales, from the molecular level to whole tissues. Understanding these emergent phenomena requires an exploration of the mapping between these different scales, and how lower-scale interactions can give rise to higher-order phenotypic structures. In this talk, I propose equipping Neural Cellular Automata (NCA) with prior knowledge of the single cell transcriptomic manifold state (by means of manifold learning) as an ideal framework for connecting these biological scales to model a specific self-organizing tissue phenotype. NCAs offer a mechanism for examining the dynamic evolution of state spaces across biological hierarchies, providing insight into how discrete components interact to yield organized, functional structures. This work aims to explore how NCAs can act as a bridge for multi-scale encodings, enabling us to formally study attractor manifolds across scales that characterize cell transcriptomic state, cell morphology state, and tissue morphology phenotypes. By drawing connections between state transition dynamics and the broader context of multi-scale topological spaces, we begin to understand how cellular interactions encode complex tissue-level phenotypes, providing a pathway towards unifying mechanistic and functional perspectives of biological systems. I invite the audience to discuss how we might further leverage the lens of topology and category theory to help formalize and takle these multi-scale systems biological challenges.
Practical information
- Informed public
- Free
Contact
- Jerome Scherer