Bayesian sample size determination for the accurate identification of the bacterial subtypes

Area of Science:

• Microbiology
• Statistical Genetics
• Bioinformatics

Background:

• Accurate sample size estimation is crucial for biosciences experiments, particularly in microbiology for resource allocation.
• Determining the optimal sampling unit for bacterial strains like E. coli presents a significant challenge.

Purpose of the Study:

• To derive a Bayesian approach with a Dirichlet prior for estimating adequate sample sizes.
• To determine E. coli subtypes efficiently using pilot study data.

Main Methods:

• Utilized five strains of E. coli, characterized by specific genes (st, eae, stx1, stx2, ial).
• Employed Polymerase Chain Reaction (PCR) on bacterial supernatants to identify gene presence.
• Applied Dirichlet Distribution as a prior within a Bayesian framework to calculate optimum sample size.

Main Results:

• Frequentist sample size calculations yielded a wide range (25 to 4475 colonies).
• The Bayesian approach, using posterior expectations, determined a more practical sample size range (291 to 443 colonies).
• PCR results showed varying prevalence of E. coli genes: eae (78.8%), stx2 (11.7%), st (3.9%), stxl2 (3.9%), and ial (1.9%).

Conclusions:

• The Bayesian approach provides an optimal sample size for E. coli subtype determination.
• This method achieves comparable statistical power to traditional techniques but with more efficient sample size calculations.
• The Bayesian method is advantageous when prior information is available, leading to better resource management.