Lineage Tracing and Predictive Computation

A tumor biopsy records the cells present at one time point, but not where those cells came from or what they will become. We develop experimental and computational methods that connect cell ancestry to molecular phenotype and treatment response. Our single-cell genetic lineage-tracing program combines static and mutable barcodes with transcriptomic, epigenetic, spatial, and genetic measurements. We use probabilistic inference and machine learning to reconstruct clonal histories, quantify cellular plasticity, and identify transitions associated with therapeutic escape.

These methods support a broader goal: predicting the composition and vulnerabilities of recurrent disease from measurements obtained earlier in a patient’s course. Software and reproducible workflows are released through the Diaz Lab GitHub organization when appropriate.