In silico design of stable single-domain antibodies with high affinity

Antibody-based therapeutics have become indispensable in modern medicine, but traditional methods of antibody discovery often present with limitations in developability, cross-reactivity, and ethical concerns. While deep learning and generative approaches have shown promise in the design of high affinity protein binders, de novo antibody design remains challenging. Here, we present EvolveX, a structure-based computational pipeline for designing antibody fragments. EvolveX utilizes ModelX and empirical force field FoldX to optimize complementarity- determining regions (CDRs) and TANGO for aggregation analysis. We demonstrate the ability of EvolveX to redesign a single-domain VHH antibody fragment targeting mouse Vsig4 to address two challenges: enhancing stability and affinity for the original target and redesigning it for high affinity to the human ortholog. The redesigned variants of VHH fragments specific to mouse Vsig4 showed improved physicochemical properties, while retaining binding affinities comparable to the original version. Notably, EvolveX improved the binding affinity of VHHs to human Vsig4 by over 1000-fold, transforming low-affinity binders into nanomolar-affinity molecules. Structural analyses by X-ray crystallography and NMR confirmed the accuracy of the designs, which display optimized interactions with the antigen. NGS and re-modelling analysis further demonstrated the efficiency of FoldX-based design pipeline. Collectively, our study highlights EvolveXs potential to overcome current limitations in antibody design, offering a powerful tool for the development of next-generation therapeutics with enhanced specificity, stability, and efficacy.