Automated mapping of multi-cellular motifs during tissue morphogenesis
Research over the last decades has identified an increasing repertoire of conserved cellular behaviors, or “motifs”, that act as building blocks of tissue morphogenesis. However, a comprehensive framework for the exploration and analysis of these motifs, similar to the frameworks used to map and discover motifs in sequence data, is yet to be established. In this talk I will present our first step towards this goal by developing a generic algorithm that can learn to recognize any sub- to multi-cellular behavior from user provided examples, and accurately map its appearances in live imaging data. Our strategy relies on the transformation of the intricate geometry, topology and molecular expression profiles of cells in a developing tissue into time-series data, thereby allowing to address the problem as a subsequence matching task. Using this approach we mapped intercalary behaviors, namely T1-transitions and rosettes, during Drosophila germband extension in wild type embryos and embryos lacking the AP patterning information, revealing differences in the kinetics of junction contraction as compared to elongation. Moreover, using Monte-Carlo simulations we show that the frequencies of T1-transitions and 5- and 6-cell rosettes can be predicted by the spatial density of contracting junctions within the tissue.
We believe that in the future our approach will begin to play in the study of tissue development the same role as standard sequence analysis is playing in the discovery of regulatory interactions in DNA and protein data.