A Probabilistic Integrated Gene Network Model of Multiple Hematopoietic Lineage Specification

Hematopoietic stem cells and their progenitor hierarchy are highly controlled by the underlying gene regulatory network. In past decades, great progress has been made in elucidating lineage-restricted transcription factors and lineage-specific gene expression patterns. With accumulation of genome-wide biological data, it is of great value to expand beyond mere transcriptional regulatory analysis to the systemic understanding. Here, we utilized a probabilistic integrated gene network for integrating heterogeneous functional genomic and proteomic data sources into predictive model of hematopoietic lineage diversification. We first constructed a naïve Bayesian network by incorporating disparate biological data, including time-series gene expression during re-dedifferentiation of leukemia along alternative paths into granulocyte or monocyte upon the treatment of differentiation-inducing agents, computationally generated transcription factor regulatory sites and microRNA targets, well-curated physical protein-protein interaction, and functional annotation data. The resultant network, coupled with binomial-based statistical analysis of the interplay between node properties and the network topology, predicted the specific hematopoietic lineage, generating testable hypotheses regarding their unified reprogramming principles. This study demonstrates the utility of the growing biological data in detailed elucidations of the orchestration of distinct hematopoietic cell fates.