Advances in Monoclonal Antibody Separation Using Physics-Informed Neural Networks

Strategic Insights -

Researchers from the Indian Institute of Technology (IIT) Patna have developed a new method for separating monoclonal antibodies from acidic charge variants using physics-informed neural networks (PINN). The method uses a combination of the Yamamoto model and Møllerup's thermodynamic approach to predict the distribution coefficient of monoclonal antibodies and their acidic variants based on salt concentration and pH. The resulting step salt gradient and pH-salt dual linear gradient elution strategy achieved high purity and yield of monoclonal antibodies.

Key Takeaways:

  • The proposed method combines the Yamamoto model and Møllerup's thermodynamic approach to predict the distribution coefficient of monoclonal antibodies and their acidic variants based on salt concentration and pH.
  • The physics-informed neural network (PINN) model was employed to identify optimal salt conditions for effective separation, achieving an R 2 score of 0.999 in just 90 seconds.
  • The estimated Gibbs free energy values were consistent with existing literature, and the predicted normalized gradient slope (GH) versus salt concentration (I) curves were within ±4.42% of experimental uncertainty.
  • The step salt gradient elution achieved 96.6% monoclonal antibody (mAb) purity and 93.1% yield, while the dual gradient elution resulted in 94% purity and 82% yield.
  • The study demonstrated that PINN modeling helped enhance chromatographic process development, requiring only three experimental data points per elution pH to effectively separate closely related impurities.
  • The research has been peer-reviewed and published in the Industrial & Engineering Chemistry Research journal.
  • The study was conducted by researchers from the Indian Institute of Technology (IIT) Patna, including Lalita Kanwar Shekhawat, Pratik Punj, Anupa Anupa, and Anurag S. Rathore.

Statistics:

  • R 2 score achieved by the PINN model: 0.999
  • Time required for PINN model prediction: 90 seconds
  • Experimental uncertainty: ±4.42%
  • Purity of monoclonal antibodies achieved by step salt gradient elution: 96.6%
  • Yield of monoclonal antibodies achieved by step salt gradient elution: 93.1%
  • Purity of monoclonal antibodies achieved by dual gradient elution: 94%
  • Yield of monoclonal antibodies achieved by dual gradient elution: 82%

Sources:

  • Physics-informed Neural Networks for Ion-exchange Chromatography-based Separation of Monoclonal Antibody and Acidic Variants. Industrial & Engineering Chemistry Research, 2025.
  • Industrial & Engineering Chemistry Research can be contacted at: Amer Chemical Soc, 1155 16TH St, NW, Washington, DC 20036, USA.
  • Lalita Kanwar Shekhawat, Indian Institute of Technology (IIT) Patna, Dept. of Chemical and Biochemical Engineering, Patna 801106, Bihar, India.