Fig. 1: A meta-atlas of the developing human neocortex.

a,b, Seven recently published transcriptomic datasets^{5,6,9,12,13,16,18} were processed through rigorous quality control metrics (nFeature_RNA > 500–1,500, min.cells = 3, percent.mt < 5) and integrated using conventional methods that identify cells that can serve as anchoring points between datasets²⁶. a, UMAPs of the resulting integrated meta-atlas show the presence of cell types and subtypes expected for the developing human cortex, and the clustering of meta-atlas cells is driven primarily by cell type identity and developmental stages⁴⁸. b, UMAPs display the normalized expression of canonical cell type markers in our integrated meta-atlas, highlighting the correspondence between cell type clusters and appropriate marker genes: HOPX (outer radial glia), EOMES (IPCs), NEUROD6 (excitatory neurons) and DLX6-AS1 (inhibitory neurons). c, Gene networks representing biological processes throughout our entire meta-atlas were identified using a meta-module strategy based on iterative, hierarchical clustering. For each of the 96 individuals in our meta-atlas, cells were first hierarchically clustered into cell types. The marker genes of these resulting clusters represented the gene expression signatures present within each individual, and we pooled the markers most representative of their assigned cluster as determined by a gene score metric rooted in specificity and enrichment. We then took this collection of cluster marker genes across all 96 individuals and conducted hierarchical clustering, binning the cluster marker genes into meta-modules. Meta-modules, therefore, comprise genes that share a similar expression pattern across all 96 individuals in our developing neocortical meta-atlas. d, Top, visual representation of the generation of meta-modules from individual cluster marker genes. The correlation between markers across all individuals was calculated based on their gene score metric across all clusters in the meta-atlas (Gene Score Correlation; green), generating a distance matrix similar to the one shown. Hierarchical clustering of markers based on their gene score correlation then binned these genes into meta-modules (purple boxes). Bottom, histogram of the number of genes represented in each of the 225 meta-modules, which range in size from 10 to 471 genes, with a median of 21 genes. mo., months; OPC, oligodendrocyte progenitor cell; scRNA-seq, single-cell RNA sequencing.