ViralMap: predicting features in viral proteins from primary sequence

Modern viral vaccines are designed to elicit an immune response against viral proteins that mediate infection, making those proteins important targets for characterization and engineering. To improve vaccine efficacy, the proteins often require changes to specific residues or domains to enhance immunogenicity and induce a protective response. These engineering strategies vary significantly across viruses, making comprehensive and accurate protein sequence annotation a crucial step for guiding vaccine design. The growing risk of novel pathogen emergence and initiatives such as the CEPI 100 Days Mission to rapidly counter "Disease X" threats heighten the need for tools that can convert viral protein sequences from newly characterized genomes or emerging variants into the annotation profiles required for antigen engineering. To address this, we developed ViralMap, a multi-label annotation model tailored for eukaryotic viral proteins. By leveraging ESM-2 language model representations, ViralMap simultaneously predicts 10 distinct annotation classes spanning domain topology and localization, post-translational modifications, and structural features directly from primary sequences. The model achieves a residue-level precision-recall area under the curve (PR-AUC) of 0.75 or greater for 7 of the 10 classes, with performance competitive with established tools across the eight benchmarked classes. Case studies on complex glycoproteins from SARS-CoV-2, HIV-1, Nipah virus, and Lassa virus demonstrate the model's ability to predict detailed residue-level annotation profiles, including for proteins from viral families not seen during training. By providing a unified, sequence-based framework for multi-label annotation, ViralMap offers a practical bridge from raw viral protein sequences to the annotation profiles required for antigen engineering.IMPORTANCEThe rapid characterization of viral proteins is critical for developing vaccines against emerging pathogens. When a new strain or virus is identified, researchers need to efficiently identify key features of these proteins to guide vaccine design. Currently, obtaining such features requires running multiple computational tools, which are generally not specialized for viruses that affect humans. ViralMap is a deep learning model that addresses this gap by predicting ten functionally relevant protein annotations simultaneously from sequence alone. Trained specifically on eukaryotic viral proteins, ViralMap aims to support the early stages of vaccine engineering for pandemic preparedness efforts.