AI-assisted design of ligands for lipocalin-2.

Introduction: Lipocalin-2 (LCN2) is an acute-phase glycoprotein whose upregulation is associated with blood-brain-barrier breakdown and neuroinflammation, making it a potential diagnostic and therapeutic target. Methods: We developed an end-to-end, AI-guided workflow to design de-novo miniproteins targeting LCN2. Backbone scaffolds were generated using RFdiffusion, sequences optimized with ProteinMPNN, and candidates filtered in silico based on AlphaFold2 confidence metrics (mean interface pAE < 10) and binding free energy predicted by Prodigy. From an initial library of 10,000 designs, five candidates were expressed and purified from E. coli. Binding affinities were assessed using biolayer interferometry (BLI), and structural interactions were analyzed via computational modeling. Results: BLI identified MiniP-2 as the lead construct, exhibiting a dissociation constant (Kd) of 4.2 nM. Structural modeling indicated that binding is primarily mediated by backbone hydrogen bonds, complemented by a stabilizing salt bridge between Arg37 of MiniP-2 and Asp97 of LCN2. Surface plasmon resonance (SPR) competition experiments demonstrated that MiniP-2 inhibits LCN2 binding to MMP-9, highlighting its potential to interfere with pathological LCN2 interactions. Discussion: These results demonstrate that a fully computational generative workflow can yield nanomolar LCN2 binders in a single design-build-test cycle. MiniP-2 represents a promising starting point for affinity maturation, structural studies, and in vivo evaluation as an imaging probe or antagonist of LCN2-mediated signaling.