As individual neurons constitute basic units of gene regulation, a major determinant of each neuron’s phenotype and function likely lies in its transcription programs. Recent studies classify neurons using high-throughput single cell RNA sequencing (scRNAseq) (Tasic et al., 2016; Zeisel et al., 2015). A major challenge is to map transcriptome-based statistical clusters, which are prone to technical noise and methodological bias, to the biological ground truth - their anatomical and physiological properties that constrain and contribute to cell function in neural circuits. In the retina, where cell types are among the best understood in the mammalian nervous system, scRNAseq has identified distinct markers that correlate to known types and suggested novel types (Shekhar et al., 2016). In the cerebral cortex, where cell type definition is often ambiguous and controversial, scRNAseq analyses have parsed multiple “transcriptional types” (Tasic et al., 2016; Zeisel et al., 2015), but their correlations to bona-fide biological types jointly defined by anatomical and physiological features remain unclear. Thus although scRNAseq allows comprehensive and quantitative measurements of gene expression, a fundamental unresolved issue is how transcription profiles might
