Inferring cell fate determinants from phenotypic expression patterns on cell lineage trees

A cell lineage tree encodes the history of where and when fate is determined in a developing cell system. It has long been appreciated that, by measuring the patterns of phenotypic expression on this lineage tree, one can infer the timing and location of important events in cell fate determination. In the simplest case, if a given cell type is restricted to one clade, the most recent common ancestor of that clade is deemed important in determining that cell fate. In practice, however, cell fates are not monophyletic and are determined by an ordered sequence of cytoplasmic and inductive factors usually occurring well before expression is observed. Inferring fate determinants from lineage tree patterns thus requires a considerably more sophisticated statistical model than simple common ancestry. Here we describe a Bayesian model that captures many of the standard patterns observed in cell lineage trees and is able to trace them back to multiple sources in the tree. Although this model has similarities with the phylogenetic comparative methods used to detect speciation events in evolutionary biology, it incorporates substantial modifications to account for the patterns of association unique to developing cell systems. The technique has been tested on published C. elegans data and applied to data from mammalian immune cells.