Graph-based tracing of dynamically functioning gene circuits in cell fate decisions with isoform resolution

Abstract

Transcription factor (TF) regulatory networks govern complex cellular processes, yet their inference from single-cell transcriptomes is hindered by TFs low abundance, tissue-specific functions, and isoform diversity. We evaluated the utility of transcription factor activity (TFA) to infer TF-TF interactions, demonstrating its superiority over expression-based methods in capturing tissue-specific, high-fidelity regulatory relationships. Using simulated and real-world transcriptomes across human hematopoiesis and early embryonic development, we validated TFAs ability to render regulatory relationships. Through utilizing the concept of propagated regulome, we uncovered TFAs utility in retrospectively resolving dynamic regulatory circuits. To address regulome limitations, we developed an embryonic stem cell-specific Consensus REgulome Database (CRED) with isoform resolution, revealing functional heterogeneity among TF isoforms. Applied to human preimplantation single-cell datasets, CRED outperformed existing regulomes, identifying isoform-specific regulators and tracing regulatory modules despite severe dropouts. Our work unleashes TFA as a robust tool for reconstructing TF networks and highlights the importance of transited, isoform-level, cell-type-specific analysis in unraveling transcriptional regulatory circuits. HighlightsO_LITFA (Transcription Factor Activity) renders regulatory relationships with high fidelity compared to expression. C_LIO_LIUsing transited regulons enables TFA to retrospectively resolve dynamic regulatory interactions in cell fate decisions. C_LIO_LIAn integrated ESC-specific transited regulome database reveals isoform-level heterogeneity. C_LIO_LITFA successfully identifies key TF regulatory modules during human preimplantation from datasets with severe dropout. C_LI