An Algorithm for Building Multi-State Classifiers Based on Collision-Induced Unfolding Data

Collision-induced unfolding (CIU) has emerged as a valuable method for distinguishing iso-cross-sectional protein ions through their distinct gas-phase unfolding trajectories. CIU shows promise as a high-throughput, structure-sensitive screening technique with potential applications in drug discovery and biotherapeutic characterization. We recently developed a CIU classification workflow to support screening applications that utilized CIU data acquired from a single protein charge state to distinguish immunoglobulin (IgG) subtypes and membrane protein lipid binding. However, distinguishing highly similar protein structures, such as those associated with biotherapeutics, can be challenging. Here, we present an expansion of this classification method that includes CIU data from multiple charge states, or indeed any perturbation to protein structure that differentially affects CIU, into a combined classifier. Using this improved method, we are able to improve the accuracy of existing, single-state classifiers for IgG subtypes and develop an activation-state-sensitive classifier for selected Src kinase inhibitors when data from a single charge state was insufficient to do so. Finally, we employ the combination of multiple charge states and stress conditions to distinguish a highly similar innovator/biosimilar biotherapeutic pair, demonstrating the potential of CIU as a rapid screening tool for drug discovery and biotherapeutic analysis.