Challenges and pitfalls of inferring microbial growth rates from lab cultures

Abstract

After more than 100 years of generating monoculture batch culture growth curves, microbial ecologists and evolutionary biologists still lack a reference method for inferring growth rates. Our work highlights the challenges of estimating the growth rate from growth curve data and shows that inaccurate estimates of growth rates significantly impact the estimated relative fitness, a principal quantity in evolution and ecology. First, we conducted a literature review and found which different types of methods are currently used to estimate growth rates. These methods differ in the meaning of the estimated growth rate parameter. Kinetic models estimate the intrinsic growth rate μ whereas statistical methods — both model-based and model-free — estimate the maximum per capita growth rate μmax. Using math and simulations, we show the conditions in which μmax is not a good estimator of μ. Then, we demonstrate that inaccurate absolute estimates of μ is not overcome by calculating relative values. Importantly, we find that poor approximations for μ sometimes lead to wrongly classifying a beneficial mutant as deleterious. Finally, we re-analyzed four published data-sets using most of the methods found by our literature review. We detected no single best-fitting model across all experiments within a data-set and found that the Gompertz models, which were among the most commonly used, were often among the worst fitting. Our study provides suggestions for how experimenters can improve their growth rate and associated relative fitness estimates and highlights a neglected but fundamental problem for nearly everyone who studies microbial populations in the lab.

Solving the stochastic dynamics of population growth

“…Importantly, our theory yields the full time-dependent probability distribution of population sizes [see equation (24)]. Based on this result, an interesting future research direction would be to improve inference methods for growth parameters, as current methods suffer from substantial limitations [49]. We propose that the present work constitutes the first step towards an exact inference method since it allows for the exact calculation of the likelihood function [89].…”

“…We consider four distinct growth models belonging to the generalized logistic growth models : Blumberg, Gompertz, Logistic, and Richards models [4]. Our choice was motivated by their widespread use to fit experimental or empirical data to estimate growth parameters in microbiology and ecological communities [38, 48, 49]. These kinetic models differ by their per capita growth rates, b N .…”

Ecological Drivers of Community Cohesion

“…Expansion of our framework to incorporate greater growth rate diversity would provide greater accuracy. However, even in a laboratory-based system, such as yeast, obtaining an accurate representation of growth rates can prove difficult ( 63 ). In complex ecosystems or when greater parameter diversity is required, it can be impossible.…”

Evolutionary rescue in a fluctuating environment

“…Mathematical growth models allow, among other things, the fitting of population dynamics data [56]. However, to date, there is no universal model that best describes any data set [41]. Our work highlights that it is crucial to correctly infer the growth type from empirical data when assessing the persistence of a population undergoing environmental change.…”

“…We also present results for the Gompertz and Richards growths, whose per capita birth rates satisfy b W,α log( K/N W ) and b W,α (1− ( N W /K ) β ), respectively (see figure 1d). These growth types, which are used to fit population growth data [40, 41], have different equilibrium sizes and per capita growth rates that may impact the probability of evolutionary rescue. The population evolves in an environment that fluctuates between two states, namely favorable F and harsh H, which impacts only the birth rate.…”