Emerging pandemics are often driven by viral spillover events. However, predicting them remains difficult due to the complex interplay between viral genetics, host adaptability, and ecological dynamics. While traditional models rely on ecological or host traits, recent advances using viral sequence data and machine learning approaches have improved prediction accuracy. However, these approaches often lack a unified framework applicable across diverse viral families and host kingdoms, and are further constrained by the limited availability of comprehensive datasets. Here, we introduce SPHAK, a simple protein-based sequence similarity framework that could quantify spillover risk and predict viral family. By focusing on proteins, SPHAK is found to be a more effective way for identifying key amino acid patterns that can distinguish viral hosts. Since proteins are directly involved in infection, they could serve as a more informative probe than genome sequences. SPHAK is both accurate and generalizable, enabling its application across diverse viral families in animal and plant hosts. Application to influenza virus sequences validates the framework’s effectiveness, with SPHAK successfully distinguishing spillover-associated protein signatures across multiple pandemic-relevant subtypes, including H1N1, H3N2, and H5N1. The predictive capacity of SPHAK supports its use in early warning systems and targeted surveillance, offering a practical tool to enhance pandemic preparedness and response.
G. et al. (Wed,) studied this question.