Discovering sparse transcription factor codes for cell states and state transitions during development

Abstract

Computational analysis of gene expression to determine both the sequence of lineage choices made by multipotent cells and to identify the genes influencing these decisions is challenging. Here we discover a pattern in the expression levels of a sparse subset of genes among cell types in B- and T-cell developmental lineages that correlates with developmental topologies. We develop a statistical framework using this pattern to simultaneously infer lineage transitions and the genes that determine these relationships. We use this technique to reconstruct the early hematopoietic and intestinal developmental trees. We extend this framework to analyze single-cell RNA-seq data from early human cortical development, inferring a neocortical-hindbrain split in early progenitor cells and the key genes that could control this lineage decision. Our work allows us to simultaneously infer both the identity and lineage of cell types as well as a small set of key genes whose expression patterns reflect these relationships.DOI: http://dx.doi.org/10.7554/eLife.20488.001

Dynamics of embryonic stem cell differentiation inferred from single-cell transcriptomics show a series of transitions through discrete cell states

“…In the accompanying paper, Furchtgott et al develop a Bayesian framework that simultaneously infers (i) cell cluster identities of the cells, $true0\left\{C\right\}\equiv \left\{c_1,c_2,\dots ,c_N\right\}$,, (ii) the sets of transitions $\left\{\text{T}\right\}$ between these clusters, (iii) the key sets of marker genes $\left\{{\alpha }_i\right\}$ that define each cell cluster and (iv) the sets of transition genes $\left\{{\beta }_i\right\}$ that define the transitions between clusters, from single-cell gene expression data $\left\{g_i\right\}$, by means of an iterative algorithm to determine the maximum likelihood estimates of these variables (Furchtgott et al, 2016). …”

“…We combined the local sets of transitions between different triplets of clusters (Figure 2—source data 2) in order to infer the most parsimonious lineage tree between the clusters (Figure 2A) (Furchtgott et al, 2016). Importantly, we obtained identical final clusters starting with different seed cluster sets using k-means clustering with the gap statistic, as well as with different threshold probability parameter values for defining transition and marker genes, showing that our results were robust to the choice of seed clusters, threshold probability value and clustering method (Figure 2—figure supplement 2A, B and C; Figure 2—figure supplement 3A and B; Materials and methods).…”

Unsupervised Trajectory Analysis of Single-Cell RNA-Seq and Imaging Data Reveals Alternative Tuft Cell Origins in the Gut

“…This problem may be somewhat alleviated in large-scale scRNA-seq studies, where cell-state transitions are driven by massive epigenetic changes reflected in gene expression programs, and genes with coordinated changes can be selected without bias. A recent Bayesian approach used the pattern of gene expression fluctuation at branch and transition points to systematically identify small sets of transcripts important for such transitions (Furchtgott et al, 2017). Future developments into unbiased feature selection from highly multiplexed gene expression data can further improve the stability of cell-state transition trajectory analysis such as p-Creode.…”

A new way to build cell lineages

“…Now, in a pair of papers in eLife, researchers at Harvard University and the Allen Institute for Brain Science report a framework that uses whole-genome mRNA expression profiling to address these questions, which they then apply to stem cell differentiation in mouse embryos (Furchtgott et al, 2017; Jang et al, 2017). The basic concept that underlies these two papers concerns the second question, which is about transitions between cell states that have already been defined in advance.…”

“…Previous attempts to address this question mostly relied on the notion that two cell states are 'close' to each other in their lineage tree if their gene expression profiles are similar (Qiu et al, 2011; Shin et al, 2015). In the first of the papers Leon Furchtgott, Samuel Melton, Vilas Menon and Sharad Ramanathan present an alternative strategy, which was motivated by an investigation of gene expression in B- and T-cells as they developed (Furchtgott et al, 2017). …”