Molecular Modeling and machine learning for predicting high-concentration antibody viscosity

High-concentration monoclonal antibody formulations are essential for subcutaneous delivery but often exhibit elevated viscosity, posing challenges in drug development, manufacturing, and administration. Understanding high concentration viscosity behavior early in development is critical for formulation optimization, yet experimental assessment remains resource- and time-intensive. In response, in-silico machine learning (ML) and molecular modeling approaches have rapidly advanced over the years to enable early-stage antibody viscosity screening. This review provides a comprehensive summary of recent ML-based advances for high-concentration antibody viscosity prediction, covering dataset generation, feature engineering, model training, validation, interpretation, and deployment. By highlighting recent advances and ongoing challenges, we aim to provide a clear roadmap for researchers interested in integrating ML and molecular modeling methods into their own antibody developability pipelines to accelerate early-stage viscosity screening and drive formulation development.