Association of wastewater determinants with fish hematological and plasma biochemical responses: Multivariate analysis approach

The aim of the study was to assess the value and applicability of multivariate tools for hematological and plasma biochemical responses of fish living in treated wastewater. Physicochemical water properties and heavy metals concentration of water in spring and fall were used as determinants of multiple fish stressors. Three methods of data analysis (Agglomerative Hierarchical Clustering, Factor Analysis, Principal Component Analysis) and one method of data modeling (Partial Least Square Regression) were applied. These methods enabled identification of clustering based on observed parameters, identification of significant variables in the observed data set, and correlation of observed variables with samples collected in different places and at different seasons. Prediction of total leukocytes, lymphocytes, granulocytes, hematocrit, glucose, alanine aminotransferase, triglycerides and cholesterol from fish blood (R-2 > 0.9) was better for fall than for spring variables, regardless the sampling site (R-2 > 0.98). For hematocrit and glucose (determination coefficient over 0.99), prediction was successful regardless the season and the sampling site. The effectiveness of prediction models was also evaluated using ratio of standard error of performance to standard deviation (RPD), and range error ratio (RER). High applicability of these models was found for multiple purposes (RPD > 8 and RER > 15), including prediction of parameters from fish blood with regard to water quality.

Automated summary

The study aimed to evaluate the value and applicability of multivariate tools for assessing hematological and plasma biochemical responses of fish living in treated wastewater. Various data analysis methods and modeling techniques were applied to identify clustering based on parameters, significant variables, and correlations between observed variables and samples collected at different places and seasons. The prediction models showed better performance in fall variables for predicting various blood parameters from fish and demonstrated high applicability for multiple purposes.