Computational Design of a De Novo Binder for Claudin-6 (CLDN6)

Claudin-6 (CLDN6) is a tight-junction membrane protein normally expressed in embryonic tissues but abnormally activated in several cancers, where it contributes to tumor growth and metastasis. Its restricted oncofetal expression makes CLDN6 an attractive target for selective cancer therapies, yet designing molecules that bind its small, flexible extracellular loops has proven challenging. Here, we report the first de novo computational design of a protein binder to our knowledge that specifically engages the extracellular domain of human CLDN6.Using RFdiffusion, we guided backbone generation around a defined CLDN6 surface epitope, performing multiple independent diffusion trajectories to explore diverse binding geometries. Candidate backbones were filtered through key metrics and visual inspection. The best-scoring backbone was then optimized using ProteinMPNN to create sequences that stabilize both the binder fold and its interface with CLDN6.The final binder was validated in silico with AlphaFold3, which predicted a well-structured CLDN6–binder complex (interface TM-score ≈ 0.73). Rosetta InterfaceAnalyzer predicted a binding free energy of –60 REU and a buried surface area of ~1200 Ų, values comparable to those of antibody–antigen complexes. Computational alanine scanning identified a cluster of key interface residues (B9–B12) that contributed most to binding. We performed in silico saturation mutagenesis and discovered that a tryptophan substitution at B12 (B12W) enhanced predicted binding energy by ~7.4 REU, increased buried surface area by +129 Ų, and strengthened hydrogen bonding by +0.85 bonds (p < 0.01).These results show us that our designed binder is able to bind tightly and form an energetically favorable interface with CLDN6. Our binder may block or occupy these regions, thereby inhibiting the oncogenic functions of CLDN6. Beyond this specific application, the study demonstrates that a workflow of integrating backbone generation, sequence design, and energetic validation can create high-affinity protein binders for membrane protein targets.