Unfolding the future: New discoveries in antibody research with the help of machine learning
MINNEAPOLIS / ST. PAUL (5/30/2025) – A new paper from the Gadkari Research Group explores novel methods in complex protein research with machine learning techniques. The paper, titled “Surface-Induced Unfolding Reveals Unique Structural Features and Enhances Machine Learning Classification Models” and authored by Rowan Matney, Gabrielle Blake, and Varun Gadkari was published in Analytical Chemistry in mid-March.
Gas-phase unfolding is a technique used to assess the structural stability of protein ions inside an ion mobility-mass spectrometer (IM-MS). It’s a fast, sensitive method that provides valuable insight into protein structure and stability. Traditionally, this is done using collision induced unfolding (CIU), where protein ions are activated by repeated collisions with neutral gas molecules, causing them to unfold. This method has been widely used to study a range of biomolecules, including antibodies, membrane proteins, and RNA. The Gadkari Research Group utilizes an alternative approach called surface induced unfolding (SIU), where ions undergo a single high-energy collision with a stainless-steel surface inside the instrument’s vacuum chamber. Their recent study shows that SIU produces unfolding data that is just as reproducible as CIU, but with greater activation efficiency, particularly useful for unfolding larger and more complex proteins like antibodies.
To handle the complexity of the data generated, the researchers integrated machine learning (ML) to classify proteins based on their unfolding profiles. IM-MS experiments generate dense, high-dimensional datasets that can be challenging and time-consuming to fully interpret using traditional analysis methods. ML provides an unbiased, data-driven approach that can capture subtle structural differences and patterns that may not be apparent through manual inspection. Using these models, the team successfully distinguished between closely related antibodies such as IgG1 and IgG4, which are over 90% identical in sequence and highly similar in structure. These antibodies are also among the most commonly engineered proteins in therapeutic development. This study demonstrates that SIU can not only match but in some cases outperform CIU, making it a powerful and efficient alternative for gas-phase protein analysis. In a broader context, this work enhances researchers’ ability to quickly and accurately characterize complex biomolecules, especially therapeutic proteins. As biologics become more central to modern medicine, techniques like SIU combined with ML can streamline drug development, improve analytical precision, and accelerate the delivery of safe, effective medical treatments.
The Gadkari Research Group – led by Assistant Professor Varun Gadkari – blends bioanalytical chemistry and chemical biology. The group’s research is aimed at developing novel mass spectrometry techniques for the structural analysis of proteins, nucleic acids, and their complexes. This work spans a range of sub-topics including biomolecular structure, neurodegenerative disease, and bioanalytical method development. Founded by Prof. Gadkari in 2022, the group is currently made up of six graduate students and one undergraduate.