Not Quite Folded: Challenges in Predicting the hNPS‐hNPSR‐Ile107 Complex With AlphaFold2 Multimer
TL;DR
The model showing higher stability in molecular dynamics simulations and consistency with known structure–activity relationships served as template to design novel hNPS analogues, which revealed the model's inability to capture hNPS bioactive conformation, as most analogues were inactive as agonists.
Abstract
The human neuropeptide S (NPS) receptor (NPSR) is a Class A peptide G protein‐coupled receptor expressed in the central nervous system and endogenously activated by NPS, a 20‐mer peptide. NPSR activation promotes cellular excitability via Gq and Gs signalling. Studies suggest that receptor antagonists may reduce drug‐seeking behaviours, whilst agonists represent innovative non‐sedating anxiolytics with memory‐enhancing effects. Despite its therapeutic potential, NPSR remains poorly characterised, with neither experimental receptor structures nor drug‐like clinical candidates available. To fill this gap, we applied a previously validated AlphaFold2 Multimer‐based protocol to model the hNPS–hNPSR complex. The model showing higher stability in molecular dynamics simulations and consistency with known structure–activity relationships served as template to design novel hNPS analogues. However, experimental validation through synthesis and in vitro pharmacological evaluation of 20 novel truncated cyclic peptides revealed the model's inability to capture hNPS bioactive conformation, as most analogues were inactive as agonists. By exploiting the stereochemical switch in hNPS hinge region, we identified four novel cyclic antagonists (17–20, pA2 in the 6.10–6.20 range). Our findings highlight strengths and limitations of current peptide‐GPCR modelling strategies and underscore the need for integrating AI predictions with experimental refinement to advance ligand discovery for challenging targets like NPSR.
