bioRxiv2025
AO_SCPLOWBSTRACTC_SCPLOWO_ST_ABSMotivationC_ST_ABSUnderstanding antigenic evolution through cross-reactivity assays is crucial for tracking viral pathogens and informing vaccine development, particularly for rapidly evolving pathogens requiring regular vaccine updates. However, existing cartography methods, commonly based on mul- tidimensional scaling (MDS), face significant challenges with sparse data, producing incomplete and inconsistent maps. There is an urgent need for robust computational methods that can accurately map antigenic relationships from incomplete experimental data while maintaining biological relevance, especially given that up to 95% of possible measurements may be missing in large-scale studies. ResultsWe present Topolow, a physics-inspired optimization framework that transforms cross- reactivity and binding affinity measurements into accurate positions in a phenotype space. By modeling antigenic relationships as a system of particles connected by springs representing measured similarities, with selective repulsion between unmeasured pairs, Topolow achieves superior performance compared to MDS. Applied to neutralization data for H3N2 influenza and HIV viruses, our method demonstrated 12% and 43% improvement in prediction accuracy respectively, while maintaining complete positioning of all antigens. Topolow determines optimal dimensionality through likelihood-based estimation, avoiding distortions due to insufficient dimensions, and demonstrates orders of magnitude better stability across multiple runs. The method effectively reduces experimental noise and bias, revealing the underlying antigenic relationships and clusters. Availability and implementationTopolow is implemented in R and freely available at [https://github.com/omid-arhami/topolow]. The package is optimized for both single machines and SLURM clusters, with parallel processing support for computational efficiency. Contactrohani@uga.edu Supplementary informationSupplementary data are available.