Obtaining maximum information from colony counts with REPOP.

REPOP accurately recovers underlying bacterial population multimodality by explicitly modeling dilution and plating uncertainties, outperforming standard multiplication methods.

• Why it matters: Understanding true bacterial heterogeneity in native environments like soil and gut microbiota is crucial, but current methods often overestimate variability and obscure subpopulations due to measurement biases.
• What they did: The authors developed REPOP, a PyTorch-based Bayesian library that reconstructs bacterial populations from plate counts, addressing complex scenarios such as multimodal distributions and biases from crowded plates, tested on simulated and real datasets.
• The result: REPOP successfully resolves distinct population peaks that naive methods miss, enabling more accurate characterization of bacterial heterogeneity and subpopulations in ecological and microbiome studies.