De novo design of two-component protein containers guided by shape

Sammanfattning

Protein containers are hollow protein assemblies with potential applications in drug delivery, nanoreactors, biosensors, and nanoparticle synthesis. Two-component containers are especially interesting because the components can be produced separately and assembled under controlled conditions. However, they are difficult to design, since both the overall container geometry and the interactions between the two components must be correct. This project investigates whether a monomer-based shape-directed design workflow can be adapted to heterodimeric building blocks. Candidate heterodimers were retrieved from the Protein Data Bank, filtered, trimmed, converted into a temporary pseudo-monomer representation, matched to native capsomer templates using Zernike shape descriptors, and redesigned in an icosahedral symmetry context. The designs were then reconstructed as explicit A/B dimers and evaluated with Rosetta and AlphaFold 3. Overall, this project shows that heterodimer-compatible shape-directed design is technically feasible, but that future work should improve input curation, template filtering, A/B dimer-interface scoring, and design optimization.