De Novo Computational Design of DARPin Binders Targeting FcεRI Alpha Chain Domain 2 via Structure-Guided Sequence Optimisation, Multi-Seed Deep Learning Prediction and Interfacial Thermodynamic Validation

Abstract

Abstract Background: Computational design of targeted protein binders against cell-surface receptorsremains a critical frontier in therapeutics and biotechnology. The high-affinity IgE receptoralpha chain (FcεRI) mediates the allergic inflammatory cascade by binding immunoglobulinE (IgE) on mast cells and basophils. Domain 2 of FcεRI (residues 89–174, PDB: 1J88)constitutes the primary IgE binding interface and represents a validated therapeutic targetfor allergic disease. Generating high-affinity synthetic protein binders against this domainwithout prior experimental co-complex structures poses a significant structural challenge. Methods: I present a fully computational pipeline combining a machine learning sequencedesign framework with multi-seed structure prediction using AlphaFold3 and Boltz-1. Designedankyrin repeat protein (DARPin) candidates were generated using a conditional variationalautoencoder trained on 172 experimentally validated DARPin–receptor complexes from theProtein Data Bank, ESM-2 protein language model embeddings, helix propensity-constrainedvariable position sampling, and iterative structure-guided interface redesign. Interface atomicinteractions, hydrogen-bonding networks, and spatial contacts were identified, extracted, andvisualized using PyMOL . Designed DARPin sequences were evaluated across five independentrandom seeds using AlphaFold3 and validated with Boltz-1. Interface thermodynamics werecharacterised using PRODIGY and PDBePISA. Results: Our top-ranked candidate (Seed 5) showed high predicted structural convergenceand interface confidence (ipTM = 0.69, minimum inter-chain pAE = 3.47 Å). Independentvalidation by Boltz-1 confirmed the predicted binding mode (ipTM = 0.61, complex pLDDT= 0.881). PDBePISA analysis revealed an extensive binding surface area of 1170.8 Å2stabilised by 15 hydrogen bonds and 8 salt bridges (∆iG = −7.6 kcal mol−1). PRODIGYpredicted binding affinity of ∆G = −12.9 kcal mol−1 (Kd ≈ 350 pM at 25◦C). The predictedinterface engages the known IgE-binding surface of FcεRI Domain 2 through 88 intermolecularcontacts including salt bridges with ASP83, GLU158 and GLU160, and hydrogen bonds withTRP84, TRP107, TRP110 and TRP153. Conclusion: These findings demonstrate a robust computational framework for de novoDARPin design against a therapeutically relevant receptor epitope, establishing a leadcandidate for experimental validation as a potential IgE-blocking therapeutic. Keywords: Protein engineering, Designed ankyrin repeat proteins (DARPins), AlphaFold3,Boltz-1, PyMOL, FcεRI, IgE receptor, protein–protein interactions, PRODIGY, PDBePISA,allergic disease.