Using Bayesian inference to characterize cell-size regulation across organisms
How organisms regulate cell size and whether this varies with growth conditions remains a fundamental question. Inferring how cells regulate their size requires separating biological control from statistical effects. Yet regulatory strategies are commonly classified by estimating the size-control parameter, , from single-cell measurements using linear regression. We show that this conventional estimator is systematically biased by measurement noise, lineage-to-lineage extrinsic variability, and finite lineage length. Consequently, the same underlying regulatory mechanism can be misclassified as sizer-, adder-, or timer-like. We develop a Bayesian framework that jointly infers and the amplitudes of intrinsic, extrinsic, and measurement noise while accounting for lineage heterogeneity and finite sampling. Across simulations spanning experimentally relevant noise levels and lineage lengths, the framework recovers and outperforms conventional regression, particularly for short and heterogeneous lineages. Reanalysis of diverse bacterial and eukaryotic datasets collected under multiple growth conditions indicates that statistical bias accounts for much of the reported variation in across organisms and growth conditions. After accounting for these biases, the datasets are broadly consistent with adder-like regulation. Thus, our results provide a general framework for distinguishing genuine biological differences from noise-induced changes in inferred regulatory strategy and revise how cell-size regulation should be inferred from lineage data.