Reconstructing differentiation networks and their regulation from time series single-cell RNA-Seq data

Article Source

Generating detailed and accurate organogenesis models using single cell RNA-seq data remains a major challenge. Current methods have relied primarily on the assumption that decedent cells are similar to their parents in terms of gene expression levels. These assumptions do not always hold for in-vivo studies which often include infrequently sampled, un-synchronized and diverse cell populations. Thus, additional information may be needed to determine the correct ordering and branching of progenitor cells and the set of transcription factors (TFs) that are active during advancing stages of organogenesis. To enable such modeling Carnegie Mellon University researchers have developed a method that learns a probabilistic model which integrates expression similarity with regulatory information to reconstruct the dynamic developmental cell trajectories. When applied to mouse lung developmental data the method accurately distinguished different cell types and lineages. Existing and new experimental data validated the ability of the method to identify key regulators of cell fate.

Learning differentiation models from single cell RNA-seq data

rna-seq

(a) Initial clusters are determined using Spectral Clustering. T0, T1, T2 represent the measurement time. (b) Initial `differentiation time’ is estimated for clusters based on difference with clusters for the first time point. DT0, DT1, DT2, DT3 denote the estimated differentiation stage. (c) Differentiation paths are constructed by connecting clusters at lower levels to their most similar parent at the level above them. (d) Regulating TFs are determined for each edge. TFs are colored based on their expression change along the edge. Increased expression: Red, Decreased expression:Blue, Stable expression:Green. Shades represent the extent of the expression change. (e) Initial model. (f) Iterating between cells and state reassignments and parameter learning until convergence. (g) Final model.

Availability – scdiff is primarily written in Python, available as an open source tool at GitHub: https://github.com/phoenixding/scdiff


Ding J, Aronow B, Kaminski N, Kitzmiller J, Whitsett J, Bar-Joseph Z. (2018) Reconstructing differentiation networks and their regulation from time series single cell expression data. Genome Res [Epub ahead of print]. [abstract]

Leave a Reply

Your email address will not be published. Required fields are marked *

*