SciPhy: A Bayesian phylogenetic framework using sequential genetic lineage tracing data
TL;DR
SciPhy is a new Bayesian phylogenetic tool that reconstructs highly accurate, time-scaled cell lineage trees and population growth dynamics from sequential CRISPR-based lineage tracing data.
Problem / question
Recent CRISPR lineage tracing technologies record the order of genetic edits over time, but existing tree-building methods like UPGMA ignore the underlying biological mechanisms of this editing process, such as variable edit rates and insertion probabilities, leading to inaccurate lineage trees and missing temporal data.
Methods
The authors developed SciPhy, a Bayesian phylogenetic framework implemented in BEAST 2, which uses a mechanistic Markov Chain Monte Carlo (MCMC) model of sequential editing. They validated SciPhy on 100 simulated datasets and benchmarked it against UPGMA and TiDeTree. They then applied it to two real DNA Typewriter datasets: a 25-day monoclonal expansion of 1,000 HEK293T cells and a 780-cell murine gastruloid derived from a single mouse embryonic stem cell (mESC).
Key findings
In simulations, SciPhy accurately recovered editing rates (Pearson's R=0.99) and outperformed UPGMA and TiDeTree in tree topology and branch length accuracy. In the HEK293T dataset, SciPhy identified highly variable insertion probabilities (e.g., 'CAT' at 16% vs. 'GCG' at less than 1%) and estimated a cell doubling time of 28 to 33 hours. In the murine gastruloid dataset, SciPhy successfully detected a significant drop in cell growth rate (from approximately 1.0 to 0.5 per day) following CHIR stimulation, matching previous imaging-based observations using only genetic barcode data.
Why it matters
SciPhy allows researchers to extract previously inaccessible developmental insights, such as absolute cell division times and dynamic population growth rates, directly from sequential single-cell lineage tracing data.
Limitations
The Bayesian inference method is computationally expensive, taking about 15 days to analyze 1,000 cells, and currently assumes equal editing rates across all sites within a tape while lacking an explicit model for missing data or tape dropout.
Takeaway
By explicitly modeling the sequential nature of CRISPR-based genetic edits, SciPhy provides a more accurate, time-calibrated reconstruction of single-cell lineage trees and developmental growth dynamics than standard distance-based clustering methods.