A comprehensive atlas of cell-type marker genes in developing human cortex via single-nucleus RNA sequencing

Identifying reliable marker genes is essential for decoding the cellular diversity and dynamic processes shaping human brain development, yet consistent markers remain lacking. We conducted a meta-analysis of single-nucleus (sn) RNA-seq datasets comprising 590,224 nuclei from the prenatal cortex and 1,150,557 nuclei from the postnatal cortex, systematically assessing marker gene expression stability, specificity, and coverage. We identified 89 high-confidence marker genes for seven prenatal brain cell types (HCP-89). Validation across independent datasets confirmed that these markers exhibit superior specificity compared to canonical genes. To further resolve astrocyte lineage cells that single markers cannot distinguish, we developed AstroRF-Net, a machine learning framework that integrates random forest and neural networks to robustly separate radial glial cells, transitional intermediates, and mature astrocytes. We also identified 259 marker genes defining postnatal brain cell types. Cross-developmental comparison revealed 164 genes with prenatal or postnatal specificity, offering insight into cell-fate regulation. These findings provide molecular and computational tools for precise cell-type annotation and future neurodevelopmental disease research.