Although we present a comprehensive analysis, our data have several limitations. First, there were technical and experimental limitations as detailed above, including doublets, sex-specific gene expression, and low-quality cells. Second, we sampled only a little more than half a million cells across the nervous system, and deeper sampling is likely to reveal additional structure that was obscured in the present study. Similarly, we used relatively shallow sequencing, and deeper sequencing using more sensitive RNA-seq methods is likely to resolve more subtypes. Third, some cell types may have been lost to differential survival or size selection biases (for example, Purkinje cells were likely undersampled here due to their size). Fourth, we have performed a very conservative clustering, designed to reveal clearly distinct major cell types, but did not analyze the substantial remaining heterogeneity within clusters. Finally, we have described only molecular cell types, but the task of linking molecular properties to functional, anatomical, morphological, and electrophysiological properties remains.
