Applied Physics Seminar | Decoding the human embryo using information theory

Abstract:
Specification of different cell types in the right locations with the right proportions is essential for tissue function. It is generally believed that cells obtain positional information from concentration gradients of molecules named morphogens, enabling them to differentiate into the right type according to their position. However, whether this model can account for patterning in the early vertebrate embryo remains unclear, largely due to challenges in both measuring and interpreting simultaneously activity of multiple morphogens and their downstream cell fate pattern. I will discuss how we have overcome these barriers in a stem cell model for human gastrulation. We developed iterative immunofluorescence staining to relate activity of six signaling pathways to twenty cell fate markers in the same cells, and applied a physics-inspired approach combining information theory and machine learning to determine quantitative relationships between position, signaling, and fate. We discovered that while positional information in individual signals is localized, the combined precision of all signals is approximately uniform in space and around three cell diameters. Strikingly, this matches the precision of cell fate patterns, suggesting a causal relationship. Moreover, while signaling distributions shift drastically in response to manipulation, the spatial precision and an empirically constructed signal-to-fate map are preserved across conditions, enabling successful prediction of fate patterning upon signaling perturbations. In summary, we discovered how information about cell fate and position is distributed across multiple signals and provide a foundation for decoding human embryonic patterning.