The cell culture medium in the channels is exchanged to L15 medium containing 10% FBS by a syringe which is plugged in each of the swabable valves
The cell culture medium in the channels is exchanged to L15 medium containing 10% FBS by a syringe which is plugged in each of the swabable valves. micropatterned single-cell arrays. We demonstrate that the integration of data from perturbation experiments allows the robust reconstruction of cell-to-cell variability, i.e., parameter densities, while each individual experiment provides insufficient information. Indeed, we show that the integration of the datasets on the population level also improves the estimates for individual cells by breaking symmetries, although each of them is only measured in one experiment. Moreover, we confirmed that the suggested approach is robust with respect to batch effects across experimental replicates and can provide mechanistic insights into the nature of batch effects. We anticipate that the proposed multi-experiment nonlinear mixed effect modeling approach will serve as a basis for the analysis of cellular heterogeneity in single-cell dynamics. Introduction Living cells show molecular and phenotypic differences at Sulbutiamine the single-cell level even in isogenic populations.1,2 Sources of cell-to-cell variability include noisy cellular processes,2 differences in cell cycle state,3 the history of individual cells,4 as well as spatio-temporal differences of the cells environment.5 Methods such as mass cytometry6 or single-cell RNA sequencing7 can provide highly multiplexed snapshots of cell-to-cell variability in thousands to millions of cells. Complementarily, time-lapse microscopy allows for the time-resolved measurement of cell-to-cell variability in the dynamic response of cells.8,9 Recently, in order to improve the high-throughput capability of single-cell time-lapse studies, microstructured arrays8,10 or microfluidic devices11 are used to restrict cells in their movement, enabling automated acquisition of single-cell fluorescence trajectories over time. Single-cell technologies already facilitated many novel insights, ranging from the analysis of population structures3,6 over the assessment of developmental trajectories12,13 to mechanistic insights into causal differences.2,14C16 To gain mechanistic insights, many studies use ordinary differential equation (ODE) models.17C20 In this spirit, earlier studies have analyzed time-lapse microscopy measurements of single-cells after transfection with synthetic mRNA to Sulbutiamine assess mRNA lifetime.21 mRNA lifetime is of fundamental interest to basic science, as it is a key parameter in many gene regulatory processes. Moreover, transient transfection of synthetic mRNA is relevant for biomedical applications, as it enables treatment of diseases via the targeted expression of proteins.22,23 Hence, a good understanding and control of the expression dynamics of therapeutic proteins is essential for treatment design.24 Yet, inference of quantitative estimates from single-cell experiments is model dependent and PRKD1 only insofar meaningful as our mechanistic understanding of many basic cellular processes, including transcription and translation, is sufficiently accurate. The model parameters can be estimated from single-cell time-lapse microscopy measurements using two different approaches: (I) The standard two-stage approach (STS) estimates single-cell parameters and population distribution parameters sequentially.25,26 First, parameters for every single cell are estimated independently by fitting an ODE to the respective trajectory. Then, a population-wide parameter distribution is reconstructed according to the single-cell parameter Sulbutiamine estimates. The STS approach enjoys great popularity,21,25C27 because it is easy to implement, as many methods and tools developed for bulk data can be applied. However, the STS approach fails to distinguish between cell-to-cell variability and uncertainty of the estimated single-cell parameters, resulting in the overestimation of cell-to-cell variability.28 This impairs applicability of the STS approach in settings with high experimental noise and sparse observations.26 (II) In contrast, the non-linear mixed effect (NLME) approach29 estimates single-cell parameters and population distribution parameters simultaneously. The single-cell parameters are considered as latent variables, which are constrained by the population distribution. The implementation of the NLME approach is more involved30C32 and its application computationally more intensive. Originally developed in pharmacology, 32 the NLME approach has recently risen in popularity for the analysis of single-cell data.25,26,33,34 It has been reported that NLME is more robust than STS in settings with huge parameter uncertainty, since it decreases doubt26,28 and eliminates estimation bias.25 The NLME approach has several advantages on the STS approach when single-cell parameters have poor practical identifiability,26,28 i.e., when the noisiness or amount of the info prohibits reliable parameter estimation. Nevertheless, structural non-identifiability35 of single-cell guidelines is difficult for the STS, aswell for the NMLE strategy. Structural non-identifiabilities, and therefore the dependable parameter estimation can be impossible because of model framework (vector field and observable), of Sulbutiamine single-cell guidelines can lead to structural non-identifiability of human population distribution guidelines36 and therefore prohibit the dependable estimation of cell-to-cell variability. For mass data, such structural non-identifiabilities could be solved by taking into consideration perturbation tests.37 For single-cell data, it really is unclear the way the thought of perturbation tests impacts non-identifiability for the NLME and STS strategy. Previous studies show how the single-cell degradation prices of mRNAs and protein are structurally non-identifiable when contemplating time-lapse microscopy.