Atsushi Mochizuki, Kyushu University
Cell lineage of a multicellular organism has been analyzed by introducing a genetic or chemical marker that is inherited from a cell to its daughter cells and is detectable even after several cell divisions. Traditionally, the intracellular marker is introduced to a single identified cell, and must be observed in many developmental stages. In this study, I introduce a new method of estimating cell lineage based on distributions of intercellular markers observed at a single stage, which, moreover, can be introduced randomly at earlier stages. Assumptions are: (1) cell lineage is invariant between embryos; (2) very few cells are marked in each experiment; and (3) the total number of replicate experiments is large. Under these assumptions, the most likely cell lineage pattern (or tree topology) can be characterized as the one that requires the least marker insertions to be compatible with the observed distributions of markers. This method is essentially the same as the principle of persimony used for ancestral phylogeny reconstruction. When the total number of cells is small, we can generate all the possible cell lineages, and then choose the one that requires the least marker insertions. If the number of cells is large, we can use clustering method in which a pair of cells with the highest correlation in marker labelling are merged sequentially. The efficiency of the clustering method in estimating the correct cell lineage is confirmed by computer simulations. The clustering method is applied to reconstruct the cell lineage of ascidian from experimental data.