SCENIC: single-cell regulatory network inference and clustering

Abstract

We present SCENIC, a computational method for simultaneous gene regulatory network reconstruction and cell-state identification from single-cell RNA-seq data (http://scenic.aertslab.org). On a compendium of single-cell data from tumors and brain, we demonstrate that cis-regulatory analysis can be exploited to guide the identification of transcription factors and cell states. SCENIC provides critical biological insights into the mechanisms driving cellular heterogeneity.

Figure 1. The SCENIC workflow and its application to the mouse brain.

(a) Co-expression modules between transcription factors and candidate target genes are inferred using GENIE3 or GRNBoost. RcisTarget identifies those modules where the regulator’s binding motif is significantly enriched across the target genes; and creates regulons with only direct targets. AUCell scores the activity of each regulon in each cell, yielding a binarized activity matrix. Cell states are based on the shared activity of regulatory subnetworks. (b) SCENIC results on the mouse brain ; cluster labels correspond to ; master regulators are color-matched with the cell types they control. (c) transcription factors confirmed by literature (A) or having brain phenotypes from MGI (B), and the enriched DNA motifs are shown. (d) t-SNE on the binary regulon activity matrix. Each cell is assigned the color of the most active GRN. (e) Accuracy of different clustering methods on this dataset.

Figure 2. Cross-species comparison of neuronal networks and cell types.

(a) DLX1/2 regulons inferred from mouse and human brain scRNA-seq data. The genes highlighted in red have associations with Dlx1/2 in GeneMANIA. (b) Reciprocal activity of human and mouse Dlx1/2 regulons on mouse and human single-cell data. (c) Joint clustering of human and mouse brain scRNA-seq data based on GRN activity. Colored TF names correspond to regulons identified both in the human and mouse SCENIC runs.

Figure 3. SCENIC overcomes tumor effects and unravels relevant cell states and GRNs in cancer.

(a-b) t-SNEs on the expression matrices, colored by tumor of origin. (c-d, f-g) t-SNEs on the binary activity matrix (e,h) after applying SCENIC. In d and g, cells are colored by GRN activity. (i) Immunohistochemistry (IHC) on 25 human melanomas using NFATC2, NFIB, ZEB1, and EPHA2 antibodies. The heatmap shows the percentage of cells that are positive for each marker in the given sample. On the right, a representative example of IHC for NFIB on a sentinel lymph node is shown (for additional images, see Supplementary Fig. 13). (j) Aggregation plots for MITF and STAT1 ChIP-seq signal on the predicted target regions, and as control randomly selected genomic regions with MITF/STAT motif occurrences.