Adopting tipping-point theory to transcriptome profiles unravels disease regulatory trajectory
Differentiation involves bifurcations between discrete cell states, each defined by a distinct gene expression profile. Single-cell RNA profiling allows the detection of bifurcations. However, while current methods capture these events, they do not identify characteristic gene signals. Here we show that BioTIP - a tipping-point theory-based analysis - can accurately, robustly, and reliably identify critical transition signals (CTSs). A CTS is a small group of genes with high covariance in expression that mark the cells approaching a bifurcation. We validated its accuracy in the cardiogenesis with known a tipping point and demonstrated the identified CTSs contain verified differentiation-driving transcription factors. We then demonstrated the application on a published mouse gastrulation dataset, validated the predicted CTSs using independent in-vivo samples, and inferred the key developing mesoderm regulator Etv2. Taken together, BioTIP is broadly applicable for the characterization of the plasticity, heterogeneity, and rapid switches in developmental processes, particularly in single-cell data analysis. HighlightsO_LIIdentifying significant critical transition signals (CTSs) from expression noise C_LIO_LIA significant CTS contains or is targeted by key transcription factors C_LIO_LIBioTIP identifies CTSs accurately and independent of trajectory topologies C_LIO_LISignificant CTSs reproducibly indicate bifurcations across datasets C_LI