Determining Cell Differentiation and Lineage based on Single Cell Entropy

Elucidating the lineages and molecular states of single cells is fundamentally important for understanding the formation and functions of complex organs such as the lung. We developed SLICE, a novel algorithm that utilizes single-cell RNA-seq (scRNA-seq) to quantitatively measure cellular differentiation states based on single cell entropy and predict cell differentiation lineages via the construction of entropy directed cell trajectories. We validated our approach using four independent scRNA-seq data sets. SLICE successfully reconstructed entropy-directed cell differentiation trajectories that have been previously experimentally validated, supporting the general applicability and high predictive accuracy of SLICE in determining cellular differentiation states and reconstructing cell differentiation lineages. In addition, we applied SLICE to scRNA-seq of embryonic mouse lung at E16.5 to identify lung mesenchymal cell lineage relationships that currently remain poorly defined. A two-branched transitional trajectory consisting of five lung fibroblastic subtypes was identified using SLICE.

The SLICE (Single Cell Lineage Inference Using Cell Expression Similarity and Entropy) algorithm consists of two major steps: (1) measuring cell differentiation states based on the calculation of single cell entropy (scEntropy) and (2) predicting cell differentiation trajectories by ordering single cells according to their scEntropy-derived differentiation states.

SLICE hypothesizes that entropy inversely correlates with cell differentiation state: high entropy is associated with higher functional uncertainty and more potential (cell stemness), while low entropy is associated with differentiated states with well-defined cell fates and more homogeneous functions. Therefore, scEntropy quantifies the differentiation state of a given cell by measuring the uncertainty in the activation of its cellular functions.

Using the differentiation states of individual cells measured by scEntropy, SLICE identifies relative stable cell states, defined as the centroids of the cells with local minimum entropies, and then predicts transitional paths following entropy reduction between stable cell states to reconstruct cell lineages.

•  Minzhe Guo*, Erik L. Bao*, Michael Wagner, Jeffrey A. Whitsett, Yan Xu. 2016. SLICE: determing cell differentiation and lineage based on single cell entropy. Nucleic Acids Research. doi:10.1093/nar/gkw1278. (* co-first author)  

Downloads:    Code and demonstrations (version a12242016)    README    

Previous Downloads
Code and demonstrations (version a12212016)    README