GPCR-CoINPocket, the computational methodology developed in ref. 1, is unaffected by the GPR37L1 cloning error and remains a powerful tool for understanding the pharmacological similarity of GPCRs solely from their sequences. CoINPocket-predicted GPCR–ligand pairings have been validated both retrospectively 1 and prospectively (for example, for GPR31 and HCA1 receptors 12), and the approach was also successfully applied to an evolutionarily unrelated lysine methyltransferase family 13. Furthermore, GPCR-CoINPocket has been used to guide GPCR homology modeling 14 and to delineate common GPCR activation mechanisms 15. However, confirming the CoINPocket ligand predictions for GPR37L1 was not possible in the absence of a validated GPR37L1 activity assay with appropriate controls. In other words, we were unable to experimentally test the hypothesized pairings. Indeed, a shortcoming in all functional GPR37L1 studies is that the limited armamentarium forces researchers to infer specificity from indirect and sometimes inconclusive data. For example, phenotypic or second messenger amplification assays with readouts many steps removed from receptor activation are vulnerable to nonspecific signal interference, as in our case for p426-r37L1 (refs. 1, 4), or to interference from the endogenous receptor context 6, 7. These challenges are common for orphan research, but the potential neuroprotective role of GPR37L1 and its links to seizures mandate that the pharmacology and physiology of GPR37L1 continue to be explored.