Statistical study of asymmetry in cell lineage data

TL;DR

This paper introduces a comprehensive statistical methodology for analyzing cell lineage data, effectively handling missing data, multiple genealogies, and dependence structures, to improve estimation accuracy and testing power.

Contribution

It presents a novel joint estimation approach using multiple datasets and models genealogies with a two-type Galton-Watson process, advancing cell lineage analysis.

Findings

Effective estimation of asymmetric bifurcating autoregressive process parameters.

Enhanced symmetry testing power through joint data analysis.

Application to E. coli data demonstrates practical utility.

Abstract

A rigorous methodology is proposed to study cell division data consisting in several observed genealogical trees of possibly different shapes. The procedure takes into account missing observations, data from different trees, as well as the dependence structure within genealogical trees. Its main new feature is the joint use of all available information from several data sets instead of single data set estimation, to avoid the drawbacks of low accuracy for estimators or low power for tests on small single-trees. The data is modeled by an asymmetric bifurcating autoregressive process and possibly missing observations are taken into account by modeling the genealogies with a two-type Galton-Watson process. Least-squares estimators of the unknown parameters of the processes are given and symmetry tests are derived. Results are applied on real data of Escherichia coli division and an empirical study of the convergence rates of the estimators and power of the tests is conducted on simulated data.