Understanding where a cell sits along developmental time is as important as identifying its type. While single-cell transcriptomics has catalogued the diversity of neural cell types, aligning them along a shared temporal axis across studies, species, and model systems remains a fundamental challenge. Here, we develop a single-cell transcriptomic 'clock' that predicts true developmental age, enabling standardized, cross-context comparisons of neural maturation. Through a meta-analysis of over 2.8 million cells from the developing human brain, we identify robust tissue-level and cell-autonomous predictors of developmental age. We find that bulk tissue composition predicts age within individual studies but lacks generalizability, whereas specific cell type proportions, particularly astrocytes and progenitors, track age reliably across studies. Using machine learning, we develop a cell type-agnostic predictor based on 462 genes that robustly tracks developmental dynamics across diverse cell types and datasets (error = 2.6 weeks). Our model accurately estimates developmental age in human neural organoids and detects disease-associated shifts. Model predictions further generalize across species, revealing 10-fold accelerated neurodevelopment in mice relative to humans. Our approach provides a robust framework to assess neural maturation across contexts, with broad relevance for developmental biology and disease modeling.
Venkatesan et al. (Wed,) studied this question.