BPS2026 – Structure-guided computational design of CXCR4-targeting single-domain antibodies

The C-X-C chemokine receptor type 4 (CXCR4) is a G protein-coupled receptor involved in immune diseases, viral infections, and cancer, making it a crucial therapeutic target. Single-domain antibodies (sdAbs) offer unique advantages for targeting GPCRs due to their small size, stability, and ability to access cryptic epitopes within transmembrane regions. To overcome limitations of traditional antibody libraries, we developed a computational approach to improve existing CXCR4-targeting sdAbs and generate novel variants by creating in silico libraries that systematically explore sequence space with property-based filtering. As a proof-of-concept template, we utilized AM3-114, a CXCR4 sdAb antagonist. Despite lacking an experimentally solved AM3-114-CXCR4 complex structure, we leveraged this established antagonist’s high-affinity binding properties to guide our design strategy. A Chai-1 predicted AM3-114-CXCR4 complex model was aligned against literature-reported critical binding residues to identify key complex contacts. Our computational pipeline integrates multiple state-of-the-art tools: RFdiffusion for initial backbone generation, ProteinMPNN for sequence optimization, and Chai-1 for structure prediction and validation. Initial designs were evaluated using Rosetta relaxation protocols and scoring filters to assess physical and chemical properties. Select designs underwent targeted mutagenesis, incorporating predicted binding residues for AM3-114-CXCR4 transmembrane pocket interactions. These rationally designed sdAbs represent systematic exploration of sequence space guided by predicted AM3-114 binding requirements and established CXCR4 interactions. SdAbs will be evaluated by biophysical properties and flow cytometry-based binding assays. Successful binders will undergo functional assays to determine agonistic versus antagonistic properties and downstream signaling effects. This work demonstrates the potential of integrating multiple computational design tools to engineer novel GPCR-targeting biologics without high-resolution structural templates, establishing a framework for rational antibody design against challenging membrane protein targets.