We first screened for gene ensembles according to GO terms, using randomized labels (AUROC~0.5) as controls. GO terms containing the keyword “synaptic” gave the highest AUROC score (0.91–0.98), suggesting that genes implicated in synaptic connectivity and function are most discriminating for PCPs (Figure 2B, Table S2). To identify more specific gene categories, we screened through all ~620 gene families annotated in the HGNC database (see Methods). We identified ~40 gene families (i.e. 7% of all gene families) with AUROC scores >0.75, generally regarded as a stringent threshold (Figure 2C, Table S3, S4). Strikingly, these gene families all fell into only 6 functional categories (Figure 2C–D): 1) cell adhesion molecules, 2) receptors for neurotransmitters and modulators, 3) voltage-gated ion channels, 4) regulatory signaling proteins, 5) neuropeptides and vesicle release machinery, 6) transcription factors. It is immediately evident from this list that except transcription factors (TFs), all gene categories encode proteins that localize along or close to cell and synaptic membrane (Figure 2D) and contribute to a singular aspect of neuronal biology - synaptic communication, which is implemented through synaptic connectivity and input-output signaling properties (Figure 2E).
