Pourhoseingholi Mohamad Amin, Dezfoulian Anahita, Nasiri Malihe, Dabiri Hosein, Zali Mohammad Reza
Research Center of Gastroenterology and Liver diseases, Shahid Beheshti University, (M.C), Tehran, Iran.
East Afr J Public Health. 2009 Apr;6 Suppl(1):37-8.
BACKGROUND & AIM: Sample size estimation is a major component of the design of virtually every experiment in biosciences. Microbiologists face a challenge when allocating resources to surveys designed to determine the sampling unit of bacterial strains of interest. In this study we derived a Bayesian approach with a dirichlet prior based on a pilot study in to find an adequate sample size for E. coli subtypes' determination.
MATERIALS & METHODS: Five strains of E. coli (st genes, eae gene, stx1 and stx2 genes, ial gene) were used in this study. The strains were grown overnight at 37 degrees C to obtain strains in a linear growth phase, according to McFarland's score. The resulting supernatants were used as templates in the PCR reactions. Then the results used with Dirichlet Distribution as a prior to find the Bayesian optimum sample size.
100 colonies were harvested from plate and examined in the PCR reaction, 50 colonies showed no specific genes, 40 detected as eae gene (78.8%), 6 colonies stx2 gene (11.7%), 2 colonies st gene (3.9%), 2 colonies stxl2 gene (3.9%) and only 1 colony detected as lall gene (1.9%). First according to the frequentist view sample size calculated indicating a different range of sample size from 25 colonies to unusual number, 4475 colonies. Then using Bayesian approach by posterior expectations instead of pilot results sample size were fund from the range of 291 colonies to 443 colonies.
The results indicated that Bayesian approach technique leads to optimal sample size with similar power in compare to traditional technique where sample size calculated without any prior information.
样本量估计几乎是生物科学中每项实验设计的主要组成部分。微生物学家在为旨在确定感兴趣细菌菌株抽样单位的调查分配资源时面临挑战。在本研究中，我们基于一项初步研究推导了一种具有狄利克雷先验的贝叶斯方法，以找到确定大肠杆菌亚型的合适样本量。
本研究使用了五株大肠杆菌（st基因、eae基因、stx1和stx2基因、ial基因）。根据麦克法兰标准，将菌株在37℃下过夜培养，以获得处于线性生长阶段的菌株。所得上清液用作PCR反应的模板。然后将结果与狄利克雷分布一起用作先验，以找到贝叶斯最优样本量。
从平板上收获100个菌落并在PCR反应中进行检测，50个菌落未显示特定基因，40个检测为eae基因（78.8%），6个菌落为stx2基因（11.7%），2个菌落为st基因（3.9%），2个菌落为stxl2基因（3.9%），只有1个菌落检测为lall基因（1.9%）。首先根据频率学派观点计算的样本量表明，样本量范围从25个菌落到不寻常的数量4475个菌落不等。然后使用贝叶斯方法，通过后验期望而不是初步结果，样本量范围为291个菌落至443个菌落。
结果表明，与在没有任何先验信息的情况下计算样本量的传统技术相比，贝叶斯方法技术能够得出具有相似效能的最优样本量。
