A computational strategy for predicting lineage specifiers in stem cell subpopulations.

Stem cell differentiation is a complex biological event. Our understanding of this process is partly hampered by the co-existence of different cell subpopulations within a given population, which are characterized by different gene expression states driven by different underlying transcriptional regulatory networks (TRNs). Such cellular heterogeneity has been recently explored with the modern single-cell gene expression profiling technologies, such as single-cell RT-PCR and RNA-seq. However, the identification of cell subpopulation-specific TRNs and genes determining specific lineage commitment (i.e., lineage specifiers) remains a challenge due to the slower development of appropriate computational and experimental workflows. Here, we propose a computational method for predicting lineage specifiers for different cell subpopulations in binary-fate differentiation events. Our method first reconstructs subpopulation-specific TRNs, which is more realistic than reconstructing a single TRN representing multiple cell subpopulations. Then, it predicts lineage specifiers based on a model that assumes that each parental stem cell subpopulation is in a stable state maintained by its specific TRN stability core. In addition, this stable state is maintained in the parental cell subpopulation by the balanced gene expression pattern of pairs of opposing lineage specifiers for mutually exclusive different daughter cell subpopulations. To this end, we devised a statistical metric for identifying opposing lineage specifier pairs that show a significant ratio change upon differentiation. Application of this computational method to three different stem cell systems predicted known and putative novel lineage specifiers, which could be experimentally tested. Our method does not require pre-selection of putative candidate genes, and can be applied to any binary-fate differentiation system for which single-cell gene expression data are available. Furthermore, this method is compatible with both single-cell RT-PCR and single-cell RNA-seq data. Given the increasing importance of single-cell gene expression data in stem cell biology and regenerative medicine, approaches like ours would be useful for the identification of lineage specifiers and their associated TRN stability cores.