Statistical recipe for quantifying microbial functional diversity from EcoPlate metabolic profiling
Statistical recipe for quantifying microbial functional diversity from EcoPlate metabolic profiling
复制标题
DOI：
10.1007/s11284-017-1554-0
复制
发表时间：
2018-01-01
影响因子：
2
通讯作者：
Matsui, Kazuaki
中科院分区：
文献类型：
--
作者：
Miki, Takeshi;Yokokawa, Taichi;Matsui, Kazuaki
EcoPlate通过监测孵育期间微孔板威尔斯孔的显色来定量微生物群落利用31种不同碳底物的能力。井的颜色模式代表代谢概况。以前的研究通常使用代表孵育最后一天三次技术重复的平均值的颜色模式，并且没有考虑底物化学多样性。然而，颜色在孵育期间波动，并且颜色在重复之间变化，破坏了区分样品之间微生物功能组成和多样性差异的统计能力。因此，我们开发了一个协议，以提高统计功效的两种方法。首先，我们优化了孵育和技术重复期间显色的数据处理。其次，我们将31个碳底物的化学结构信息纳入计算。我们的框架在R环境中实现为协议，能够比较不同计算方法之间的统计功效。当我们将其应用于水生微宇宙和森林土壤系统的数据时，我们观察到当我们在孵育过程中结合时间模式而不是仅使用端点数据时，统计功效有了实质性的提高。使用技术重复的最大值或最小值有时也比平均值更好。基于模糊集理论的化学结构信息可以提高统计功效，但仅当考虑相对颜色密度信息时;当图案首次二值化为代谢活性的存在或不存在时，没有看到。最后，我们讨论了改进这些方法的研究方向，并为将我们的方法应用于其他数据集提供了一些实际考虑。
EcoPlate quantifies the ability of a microbial community to utilize 31 distinct carbon substrates, by monitoring color development of microplate wells during incubation. Well color patterns represent metabolic profiles. Previous studies typically used color patterns representing average values of three technical replicates on the final day of the incubation and did not consider substrate chemical diversity. However, color fluctuates during incubation and color varies between replicates, undermining statistical power to distinguish differences among samples in microbial functional composition and diversity. Therefore, we developed a protocol to improve statistical power with two approaches. First, we optimized data treatment for color development during incubation and technical replicates. Second, we incorporated chemical structural information for the 31 carbon substrates into the computation. Our framework implemented as the protocol in the R environment is able to compare the statistical power among different calculation methods. When we applied it to data from aquatic microcosm and forest soil systems, we observed substantial improvement in statistical power when we incorporated temporal patterns during incubation instead of using only endpoint data. Using maximum or minimum values of technical replicates also sometimes gave better results than averages. Incorporating chemical structural information based on fuzzy set theory could improve statistical power but only when relative color density information was considered; it was not seen when the pattern was first binarized into the presence or absence of metabolic activity. Finally, we discuss research directions to improve these approaches and offer some practical considerations for applying our methods to other datasets.