Elucidating Protein Quaternary Structure in Near-Native Solution: An AI-Assisted Hydrodynamic Approach

Elucidating the quaternary structures of protein complexes under physiological conditions remains a central challenge for structural biology. Here, we introduce an integrative analytical pipeline that combines size exclusion chromatography (SEC), Taylor dispersion analysis (TDA), native mass spectrometry (nMS), and artificial intelligence to resolve macromolecular architectures under near-native solution conditions. The workflow begins with the SEC-based separation of heterogeneous mixtures, followed by online TDA in a PEEK capillary to probe hydrodynamic behavior via laminar-flow-induced dispersion. By deconvolving pre- and postdispersion chromatograms, a system transfer function is derived and subjected to Gaussian fitting to yield precise measurements of hydrodynamic radii. These experimentally derived radii are subsequently applied as biophysical constraints in an AlphaFold 3-based structural selection process. We validated this method using a six-protein model mixture, demonstrating its ability to simultaneously determine the hydrodynamic radii of individual components within complex mixtures. The methodology was further applied to three protein complexes (FCGRT-B2M, streptavidin, and concanavalin A), allowing for the precise measurement of their oligomeric hydrodynamic radii and the determination of their quaternary structures. By bridging high-resolution analytical chemistry with AI-driven structural prediction, this strategy offers a robust experimentally grounded framework for characterizing macromolecular assemblies.