Lineage EM algorithm for inferring latent states from cellular lineage trees.

Area of Science:

• Cell biology
• Microfluidics
• Computational biology

Background:

• Phenotypic variability in cell populations, like bacterial persistence, serves as bet-hedging against environmental changes.
• Understanding cell phenotype and inheritance is crucial for controlling these strategies.
• Current microfluidic lineage data is insufficient for direct phenotype identification.

Purpose of the Study:

• To clarify how survivorship bias distorts statistical estimations from lineage data.
• To propose a novel algorithm for latent-variable estimation that overcomes survivorship bias.
• To provide a statistical method for identifying cell traits from various lineage data.

Main Methods:

• Clarification of survivorship bias effects on statistical estimations.
• Development of a latent-variable estimation algorithm based on the expectation-maximization (EM) algorithm.
• Implementation of the lineage EM algorithm (LEM) to address bias in lineage tree analysis.

Main Results:

• Demonstration of how survivorship bias impacts statistical inferences from cell lineage data.
• Introduction of the lineage EM algorithm (LEM) to correct for survivorship bias.
• Development of a robust statistical framework for phenotype inference from lineage data.

Conclusions:

• The lineage EM algorithm (LEM) effectively resolves survivorship bias in inferring cell phenotypes from lineage data.
• LEM offers a versatile statistical approach applicable to diverse lineage tracing datasets.
• This work advances the understanding and control of cellular bet-hedging strategies.