Significant cell-to-cell heterogeneity is ubiquitously-observed in isogenic cell populations. Cells respond differently to a same stimulation. For example, accounting for such heterogeneity is essential to quantitatively understand why some bacteria survive antibiotic treatments, some cancer cells escape drug-induced suicide, or some cells are not infected by pathogens. In a synthetic biology context, accounting for cellular heterogeneities is also essential to engineer cells performing robustly a desired function. The origins of the variability of biological processes are multifarious. From systems and synthetic biology perspectives, understanding the exact contributions of the different sources of heterogeneity on the variability of cell responses is a central question.
Consequently, there have been significant efforts to develop experimental methods allowing for the observation of cellular processes at the single-cell level rather than at the population-averaged levels as it is traditionally the case. Comparatively, the development of single-cell models is still lagging behind. Indeed, current dynamical models of biological processes treat all cells as identical, either explicitly by purely neglecting variability (standard ordinary differential equation models) or implicitly by representing all variability as coming from different outcomes of a same stochastic process (standard continuous time Markov chain models). These two standard frameworks adopt a “mean-cell” view, in which all cells have identical, “mean” parameter values.
Unfortunately, neglecting parameter differences between cells raises fundamental issues. Indeed, for non-linear dynamical systems, the mean behaviour of a population of individuals (cells here) is different from the behaviour of the “mean individual” (the individual having mean parameters, if it exists), and the former should not be used for parameter estimation. One should therefore adopt a framework in which each cell is described by individual parameters (single-cell models), and the whole cell population is described by distributions of parameters (population models).
The identification of models with (multidimensional) parameter distributions is statistically challenging. Not only means and variances but also correlations between parameters need to be estimated. The problem of inferring quantitative models capturing cell-to-cell differences in parameter values from experimental data has barely been addressed. In this project, we will leverage and extend the mixed-effects (ME) modelling framework and identification tools.
Attributing individual parameters to individual cells also opens novel perspectives. For example by analysing the different responses of cells subjected to a same stimulation in connection with their local environments, one can learn how environmental factors influence intracellular processes. Going further, once equipped with the proper statistical tools for the calibration of single-cell models, one can take advantage on recent technological methods that send distinct stimulations to each individual cell, to parallelize experiments at the single-cell level so as to maximize the information content of the experiments.
In this project, we will use the ME modelling framework for the calibration of population and single-cell models to single-cell longitudinal observations. Going beyond the mere application of the standard theory, we will exploit family relations to better constrain individual parameters and investigate the inclusion of slow temporal changes of parameters in the identification framework. We will also develop optimal experimental design (OED) methods for the identification of ME models. These methods will be implemented and benchmarked on real-systems via their deployment on an innovative experimental platform. This work will be preceded by an in-depth analysis of the identifiability of our ME models.
Monsieur Gregory Batt (Centre de recherche Inria Saclay - Ile-de-France - Equipe projet LIFEWARE)
The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.
IBIS Centre de recherche Inria Grenoble - Rhône-Alpes - Equipe projet IBIS
MSC Laboratoire Matière et Systèmes Complexes (UMR 7057; CNRS and Paris Diderot Univ.)
XPOP Centre de recherche Inria Saclay - Ile-de-France - Equipe projet XPOP
LIFEWARE Centre de recherche Inria Saclay - Ile-de-France - Equipe projet LIFEWARE
Help of the ANR 493,199 euros
Beginning and duration of the scientific project: December 2016 - 48 Months