Computational design and functional validation of peptide modulators for AMPAR and NMDAR

Abstract Ionotropic glutamate receptors AMPAR and NMDAR are central to synaptic transmission and key targets in neurological disease therapy. Here, we developed a de novo design pipeline combining a generative neural network, AlphaFold2 structural predictions, and molecular dynamics to create short, high-affinity, receptor-selective peptide modulators. The algorithm targets receptor hotspots, optimizing for minimal contact distance and high structural confidence, and avoids mode collapse through position-specific scoring matrix integration. In silico evaluation of top candidates exhibited strong binding affinity, low RMSD fluctuations, and stable hydrogen bonding. Selectivity tests confirmed minimal cross-binding between AMPAR- and NMDAR-specific peptides. Top candidates were synthesized and evaluated in in vitro experiments. In rat cortical and hippocampal cultures, peptides induced robust, concentration-dependent intracellular Ca2+ influx, abolished by extracellular calcium removal or selective antagonists (CNQX, MK-801), thus confirming the positive modulatory action of designed peptides. Cross-inhibition assays demonstrated strict receptor specificity. Although calcium responses were lower than those elicited by conventional ligands (AMPA and NMDA), functional modulation was confirmed. This integrated computational–experimental framework enables efficient development of receptor-specific peptide modulators and offers a promising strategy for therapeutic discovery.