No Chains Attached: Tree-Free Molecular Clock Dating via Distance Geometry and Evolutionary Foundation Models

Molecular clock dating translates pathogen sequence divergence into calendar time. Standard Bayesian phylodynamic methods characterize evolutionary rates with exceptional fidelity, but navigating combinatorial tree space imposes heavy computational burdens and forces aggressive sequence subsampling. We present ChronAeon, a tree-free framework executing in seconds on commodity hardware. By pairing continuous distance geometry with cross-taxa attention correlation kernels from an evolutionary foundation transformer, HyphAeon, the framework estimates substitution rates, ancestral root horizons, exact analytical confidence intervals, and co-circulating sublineage clocks without reconstructing trees. Across 42 published empirical benchmarks (N = 29 to 1,610 sequences), the framework recapitulates published Bayesian substitution rates and root emergence dates. On 280 synthetic benchmark alignments, tree-free regression maintains 93%–98% confidence interval coverage under branch rate variation where existing heuristic methods experience total interval collapse. Systematically mapping operational boundaries across 124,349 forward simulations and an eight-dimensional parameter sweep, we show that standard birth-death diversification with longitudinal sampling yields sub-year root date precision (< 7% relative error). Acute low divergence and heavy surveillance clustering expand error margins, which automated shrinkage safeguards effectively bound. Existing phylogenetic pipelines struggle to process massive, uncurated archives containing mixed divergent subtypes. Ingesting a century-long census of 49,875 Influenza A genomes across 16 divergent subtypes, the algorithm isolates 453 temporal outliers and partitions 72 distinct sublineage clocks in minutes without manual curation. Available as an open-source Python package and a zero-install private browser application, ChronAeon provides an accessible computational foundation for real-time pathogen genomic surveillance.