# Assessing methods for estimating microbial lag phase duration: a comparative analysis using Saccharomyces cerevisiae empirical and simulated data

## TL;DR

This paper compares methods for estimating the lag phase in microbial growth, using both real and simulated data from yeast to assess accuracy and reliability under different conditions.

## Contribution

The study evaluates how different assumptions and data characteristics affect lag phase estimation methods, providing insights into their performance under various biological and measurement conditions.

## Key findings

- Infrequent measurements and noisy data increase bias and variance in lag phase estimation.
- Model-fitting methods perform best when dealing with noisy data.
- Biological assumptions about cell division during lag phase significantly impact estimation accuracy.

## Abstract

The lag phase is a temporary, nonreplicative period observed when a microbial population is introduced to a new, nutrient-rich environment. Although the theoretical concept of growth phases is clear, the practical application of methods for estimating lag lengths is often challenging. In fact, there are two distinct assumptions: (i) that cells do not divide at all during the lag phase or (ii) that they divide but at a suboptimal rate. Therefore, the choice of method should consider not only technical limitations but also consistency with the biological context. Here, we investigate the performance of the most common lag estimation methods, using empirical and simulated datasets. We apply different biological scenarios and simulate curves with varying parameters (i.e. growth rate, noise level, and frequency of measurements) to test their impact on the estimated lag phase duration. Our validation shows that infrequent measurements, low growth rate, longer lag phases, or higher level of noise in the measurements result in higher bias and higher variance of lag estimation. Additionally, in case of noisy data, the methods relying on model fitting perform best.

The effect of method assumptions (i.e. no cell divisions or divisions at a suboptimal rate) on the estimated lag phase lengths was investigated using empirical and simulated growth curves.