Lastly, we used TyCHE to reconstruct the evolution of a hypermutating glioma tumor lineage and bacterial lung infection. In both cases, existing strict and relaxed clocks gave unrealistic trees but TyCHE's matched well with patient history. (6/7)

We then used TyCHE to analyze longitudinal BCRs from influenza vaccinees. TyCHE correctly inferred cycles of B cell evolution from primary GC to memory B cells to recall GCs, while existing models did not. It even inferred unsampled GC reactions at the start of each lineage! (5/7)

When applied to BCRs from HIV-infected donors, TyCHE correctly inferred the direction and patterns of germinal center to memory B cell formation. Existing models did not. Strikingly, TyCHE's time trees matched closely with clinically-estimated dates of infection. (4/7)

To validate TyCHE, we developed SimBLE, which simulates DNA sequences evolving under type-linked evolution, particularly B cells in germinal center reactions. We show that SimBLE simulates accurate BCRs, and that TyCHE outperforms existing clock models in BEAST (3/7)

TyCHE estimates trees, node dates, and cell types while allowing each subpopoulation (e.g. germinal center vs memory B cells) to evolve at a different rate. Because it's a BEAST2 package, TyCHE estimates all parameters using Bayesian MCMC. (2/7)