School of Biological Sciences, 1-33 Chlorine Gardens, Belfast BT9 5AJ, United Kingdom of Great Britain and Northern Ireland.
School of Natural and Built Environment, Elmwood Avenue, Belfast BT7 1NN, United Kingdom of Great Britain and Northern Ireland.
Sci Total Environ. 2021 May 15;769:145246. doi: 10.1016/j.scitotenv.2021.145246. Epub 2021 Jan 20.
Potentially Toxic Elements (PTEs) otherwise known as heavy metals are ubiquitous in soils and can have a range of negative health and environmental impacts. In terrestrial systems understanding how PTEs move in the environment is made challenging by the complex interactions within soil and the wider environment and the compositional nature of PTEs. PTEs are compositional because data of individual PTEs within in a sample are ratios which may be under a sum constraint, where individual components sum up to a whole. In this study three different scenarios were considered, one using the centred log ratio transformation (clr) a compositional transformation, the more "traditional" log10 transformation (log10) and untransformed data acting as a comparison (unt) were applied to four different datasets. Three were the Liver, Muscle and Kidney tissue of Eurasian Badgers (Meles meles) and the fourth was soil and data were extracted from a regional geospatial survey. Cluster analysis demonstrated that the clr and log10 transformation were able to resolve compositional trends at the point of the individual sample, whilst unt could not and did not meet the preconditions for the next phase of analysis. At the level of compositional trends between PTEs complex heatmaps demonstrated that clr was able to isolate PTE relationships and highlight commonalities between different datasets, whilst log10 could not. In the final phase, principal component analysis (PCA) of the clr transformation showed similarities between the signals in the soft tissues and the disparities they had with soil, whilst the log10 transformation was unable to achieve this. Overall, the clr transformation was shown to perform more consistently under a variety of analytical scenarios and the compositional approach will provide more realistic interpretations about PTEs in both soil and animal soft tissue than the log10 or unt conditions.
潜在有毒元素（PTEs），也称为重金属，在土壤中普遍存在，可能对健康和环境造成一系列负面影响。在陆地系统中，由于土壤和更广泛环境内部的复杂相互作用以及 PTEs 的组成性质，了解 PTEs 在环境中的迁移方式具有挑战性。PTEs 是组成性的，因为样品中单个 PTE 的数据是比值，这些比值可能受到和约束，其中各个成分加起来是一个整体。在这项研究中，考虑了三种不同情况，一种是使用中心对数比变换（clr），一种组成性变换，更“传统”的对数变换（log10）和未变换的数据作为比较（unt），应用于四个不同数据集。三个是欧亚獾（Meles meles）的肝脏、肌肉和肾脏组织，第四个是土壤，数据来自区域地理空间调查。聚类分析表明，clr 和 log10 变换能够在单个样本点解析组成趋势，而 unt 则不能，也不符合下一阶段分析的前提条件。在 PTE 之间的组成趋势水平上，复杂的热图表明 clr 能够分离 PTE 关系，并突出不同数据集之间的共性，而 log10 则不能。在最后阶段，clr 变换的主成分分析（PCA）显示软组织结构中的信号与土壤之间存在相似性，而 log10 变换则无法实现这一点。总体而言，clr 变换在各种分析情况下表现更为一致，组成方法将比 log10 或 unt 条件更真实地解释土壤和动物软组织中的 PTEs。
