SCENIC is a computational method for simultaneously reconstructing gene regulatory networks (GRNs) and identifying stable cell states from single-cell RNA-seq data. It outperforms existing approaches in cell clustering and transcription factor identification, and is robust to batch effects and technical biases. SCENIC was applied to mouse and human brain data, revealing gene regulatory networks underlying distinct cell states. In melanoma, SCENIC identified a proliferative state driven by MITF and STAT, and an invasive state governed by NFATC2 and NFIB. It also validated predictions by showing that two transcription factors are predominantly expressed in early metastatic sentinel lymph nodes. SCENIC is a network-centric approach that allows tracing genomic regulatory programs and mapping cellular identities. It is available as an R workflow using three new R/Bioconductor packages: GENIE3, RcisTarget, and AUCell. SCENIC is scalable and flexible, enabling analysis of large datasets. It was tested on various datasets, including mouse and human brain data, and melanoma data. SCENIC identified distinct cell states in the mouse brain, including interneurons, and validated cross-species regulatory networks. It also identified regulatory networks in human melanoma, revealing a proliferative and invasive state. SCENIC outperformed standard clustering methods in identifying cell types and was robust to normalization and batch effects. It was applied to large datasets, including 1.3 million cells from the embryonic mouse brain. SCENIC identified gene regulatory networks in cancer, revealing distinct cell states in oligodendroglioma and melanoma. It identified a cycling cell state in both cancers, and validated the role of NFATC2 in melanoma progression. SCENIC provides a general method for analyzing single-cell RNA-seq data, using transcription factors and cis-regulatory sequences to guide the discovery of cellular states. It is a powerful tool for understanding cellular heterogeneity and gene regulatory networks in complex biological samples.SCENIC is a computational method for simultaneously reconstructing gene regulatory networks (GRNs) and identifying stable cell states from single-cell RNA-seq data. It outperforms existing approaches in cell clustering and transcription factor identification, and is robust to batch effects and technical biases. SCENIC was applied to mouse and human brain data, revealing gene regulatory networks underlying distinct cell states. In melanoma, SCENIC identified a proliferative state driven by MITF and STAT, and an invasive state governed by NFATC2 and NFIB. It also validated predictions by showing that two transcription factors are predominantly expressed in early metastatic sentinel lymph nodes. SCENIC is a network-centric approach that allows tracing genomic regulatory programs and mapping cellular identities. It is available as an R workflow using three new R/Bioconductor packages: GENIE3, RcisTarget, and AUCell. SCENIC is scalable and flexible, enabling analysis of large datasets. It was tested on various datasets, including mouse and human brain data, and melanoma data. SCENIC identified distinct cell states in the mouse brain, including interneurons, and validated cross-species regulatory networks. It also identified regulatory networks in human melanoma, revealing a proliferative and invasive state. SCENIC outperformed standard clustering methods in identifying cell types and was robust to normalization and batch effects. It was applied to large datasets, including 1.3 million cells from the embryonic mouse brain. SCENIC identified gene regulatory networks in cancer, revealing distinct cell states in oligodendroglioma and melanoma. It identified a cycling cell state in both cancers, and validated the role of NFATC2 in melanoma progression. SCENIC provides a general method for analyzing single-cell RNA-seq data, using transcription factors and cis-regulatory sequences to guide the discovery of cellular states. It is a powerful tool for understanding cellular heterogeneity and gene regulatory networks in complex biological samples.