Single-cell entropy for accurate estimation of differentiation potency from a cell’s transcriptome

Abstract

The ability to quantify differentiation potential of single cells is a task of critical importance. Here we demonstrate, using over 7,000 single-cell RNA-Seq profiles, that differentiation potency of a single cell can be approximated by computing the signalling promiscuity, or entropy, of a cell's transcriptome in the context of an interaction network, without the need for feature selection. We show that signalling entropy provides a more accurate and robust potency estimate than other entropy-based measures, driven in part by a subtle positive correlation between the transcriptome and connectome. Signalling entropy identifies known cell subpopulations of varying potency and drug resistant cancer stem-cell phenotypes, including those derived from circulating tumour cells. It further reveals that expression heterogeneity within single-cell populations is regulated. In summary, signalling entropy allows in silico estimation of the differentiation potency and plasticity of single cells and bulk samples, providing a means to identify normal and cancer stem-cell phenotypes.

Single-cell transcriptional diversity is a hallmark of developmental potential

“…Moreover, it remains difficult to distinguish quiescent (non-cycling) adult stem cells with longterm regenerative potential from more specialized cells using existing in silico approaches. While gene expression-based models can potentially overcome these limitations (e.g., transcriptional entropy [18][19][20] , pluripotency-associated gene sets 21 and machine learning strategies 22 ), their relative utility across diverse developmental systems and single-cell sequencing technologies is still unclear.…”

“…We sought to identify robust, RNA-based determinants of developmental potential without the need for a priori knowledge of developmental direction or intermediate cell states marking cell fate transitions. Toward this end, we evaluated ~19,000 potential correlates of cell potency in scRNA-seq data, including all available gene sets in the Molecular Signatures Database (n = 17,810) 23 , 896 gene sets covering transcription factor binding sites from ENCODE 24 and ChEA 25 , an mRNA-expression-derived stemness index (mRNAsi) 22 , and three computational techniques that infer stemness as a measure of transcriptional entropy (StemID, SCENT, SLICE [18][19][20]. We also explored the utility of 'gene counts', or the number of detectably expressed genes per cell, which has been anecdotally observed to correlate with differentiation status 26-28 , but not yet comprehensively evaluated (Methods).…”

“…To assess these features, we compiled a training cohort consisting of nine gold standard scRNA-seq datasets with experimentally-confirmed differentiation trajectories. These datasets were selected to prioritize commonly used benchmarking datasets from prior studies 12,[18][19][20]26,29 and to ensure a broad sampling of unique developmental states from the mammalian zygote to terminally differentiated cells 26,30 . Overall, the training cohort encompassed 3,174 single cells spanning 49 phenotypes, six tissue types, and three scRNA-seq platforms ( Fig.…”

Defining epidermal basal cell states during skin homeostasis and wound healing using single-cell transcriptomics

“…Single cell Energies (scEnergy) were calculated, projected onto UMAP space, and used to infer cell state transition probabilities and lineage relationships. We found that lower energies, while typically associated with committed/differentiated cell states (Jin et al, 2018;Teschendorff and Enver, 2017), are also associated with a quiescent cell state because the known quiescent Bu-HFSC (CD34 + ) population showed the lowest scEnergy of all the skin epithelial cell types ( Figure 3D, 4E, and S4H; see Methods section). Using increasingly robust parameters, which incorporate increasing numbers of UW epidermal cells from each cell state to infer lineage progression, scEpath predicted a near-linear path that originates from the Col17a1 Hi basal cell state that displayed the lowest energy of all interfollicular epidermal cells ( Figure 6E, S12C, and S12G).…”

Single cell transcriptomics of human epidermis reveals basal stem cell transition states

“…Previous studies have suggested that in silico differentiation potency and plasticity of single cells can be approximated by computing the signaling promiscuity of a cell's transcriptome (Teschendorff and Enver, 2017). We developed a similar algorithm to calculate the Cellular Entropy (ξ) of single cells called Cellular Entropy Estimator (CEE) -now a feature that has been incorporated into SoptSC (see Methods).…”