Motivation <p>VaxiGen is the successor of VaxiJen, the first server for alignment-independent prediction of protective antigens. Unlike traditional tools that rely on sequence alignment, VaxiJen was designed to classify antigens solely based on their physicochemical properties. Since its launch in 2007, a wealth of new experimentally validated data on protective immunogens from diverse sources has become available, creating the need for an updated approach. The present study addresses this need by developing new machine learning models for predicting the immunogenicity of antigens of bacterial, viral, and tumor origin.</p> Results <p>Bacterial and viral protein datasets were employed to construct models for predicting whole-protein immunogenicity. For tumor immunogenicity, two model types were developed: one trained on whole proteins and another on tumor-derived peptides. The performance of all models was assessed by 10-fold cross-validation and external test sets. The resulting models have been integrated into the VaxiGen web server. VaxiGen offers improved prediction of bacterial, viral, and tumor antigens while maintaining an open and extensible framework that supports the integration of future models and the refinement of existing ones. The server can be utilized as a standalone tool or in combination with alignment-based prediction methods.</p> Availability <p>The VaxiGen web server is freely accessible at <a href="https://www.ddg-pharmfac.net/vaxigen/">https://www.ddg-pharmfac.net/vaxigen/</a>.</p>