REPOP: bacterial population quantification from plate counts.

Area of Science:

• Microbiology
• Computational Biology
• Statistical Modeling

Background:

• Bacterial counts from natural environments exhibit high variability, often misattributed to biological factors.
• Standard dilution and plating methods overestimate heterogeneity by conflating subsampling uncertainty with biological variance.
• This overestimation can obscure genuine subpopulation structures, as seen in C. elegans gut microbiota studies.

Purpose of the Study:

• To develop a method that explicitly accounts for uncertainty introduced by dilution and plating randomness in bacterial population estimation.
• To introduce REPOP, a PyTorch-based library for reconstructing bacterial populations from plate data within a Bayesian framework.
• To address complex scenarios, including multimodal populations and bias from excluding crowded plates.

Main Methods:

• A Bayesian framework implemented in a PyTorch-based library named REPOP.
• Accounting for uncertainty from dilution and plating randomness.
• Correcting for bias introduced by excluding overly crowded plates.

Main Results:

• REPOP resolves distinct population peaks obscured by standard multiplication methods.
• The library accurately recovers underlying multimodality in bacterial populations by properly accounting for error propagation.
• Demonstrated success on simulated datasets and experimental data, including C. elegans gut microbiota.

Conclusions:

• REPOP provides a more accurate estimation of bacterial population structures by addressing measurement uncertainties.
• The library overcomes limitations of traditional methods, revealing true heterogeneity and subpopulations.
• REPOP is a valuable tool for analyzing microbial communities and other count-based biological data.