Individual cells are phenotypically heterogeneous even with the same genetic background. The heterogeneity endows the cells with different growth capabilities. This difference in turn works as the driving force of the natural selection among the cells, and the population gradually adapt to the environment as the population level.
Therefore, quantification of the growth capabilities from data is crucial for understanding how the evolutionary learning of the cells is shaped and regulated in a population.
To this end, we propose an Expectation-Maximization (EM) based algorithm to infer the growth capability as hidden states of cells from cellular lineage trees. We modified the EM algorithm to account for the tree structure of the lineage tree and resolved the sampling bias problem (survivorship bias) that inevitably appears in an evolving population.
By applying our method to a lineage tree data of E.coli, we identified hidden states of the cells that may be linked to the slowing changing activity of the cells. This method may contribute to facilitating our quantitative understanding on the Darwinian adaptation operating at the cellular level.