Gene Expression Trees in Blood Cell Development

We present a novel statistical framework for analyzing transcription during cell development. In particular, we focus on data from mouse lymphoid lineages [1], which is part of blood cell development. Gene expression data of cells of several distinguishable developmental stages fosters the elucidation of the underlying molecular processes, which change gradually over time and lock cells in certain lineages. Large-scale analysis of this data requires a computational framework for tasks ranging from visualization, querying, and finding clusters of similar genes, to answering detailed questions about the functional roles of individual genes and their similarities and differences. In development, we have temporal sequences, in which one cell type undergoes a given differentiation process and turns into a new cell type. Furthermore, there can be branching points at which cells at particular developmental stages differentiate into two or more cell types, and subsequently follow different differentiation paths. To address the tree-like structure of these processes, we choose to represent mRNA expression data during cell development and differentiation with tree models [2], which model dependencies during differentiation, and to combine several of these models in a mixture. Together, we obtain a robust and flexible statistical model for analyzing and clustering genome-wide mRNA expression data sets, in which the inherent dependencies between stages can be seen and overlapping clusters are allowed. Additionally, we combine sequence information, as given by finding microRNA binding sites [3], and genes with similar expression profiles, as given by the clustering results, to perform a more detailed microRNA target prediction (see Figure 1).