Associação entre indicadores ecotoxicológicos e de poluição do sedimento
Data
2023-11-22
Autores
Fonseca, Marina Ferrel 
Orientadores
Fonseca, Gustavo Fernandes Camargo
Tipo
Dissertação de mestrado
Título da Revista
ISSN da Revista
Título de Volume
Resumo
Neste estudo, a associação entre a concentração de metais, toxicidade em organismos alvo e biomarcadores foi investigada. O objetivo foi testar se indicadores biológicos (biomarcadores, bioacumulação e toxicidade) podem prever as concentrações de poluentes no sedimento. Cinco estudos já publicados formaram o conjunto de dados. Para as análises, somente as variáveis comuns a todos os estudos foram consideradas. Ao todo foram 9 variáveis do sedimento e 15 indicadores biológicos. Os dados foram analisados por meio de algoritmos de aprendizado de máquinas. 80% dos dados foram utilizados para treinar o modelo e 20% para teste. Após a padronização, os dados de contaminação de todos os estudos foram agrupados em cinco conjuntos. Os resultados das análises de correlação mostraram a ausência de correlação entre os indicadores biológicos evidenciando a importância de analisá-los em conjunto. Na fase de treino, o modelo de aprendizado de máquina selecionou um sub-conjunto de indicadores biológicos para prever o gradiente de poluição do sedimento, obtendo 91,479% de acurácia. No conjunto de teste esta acurácia subiu para 100%. A toxicidade ocorreu principalmente nos grupos com maior concentração de metais, sugerindo uma relação entre os contaminantes e os efeitos nos organismos. Dentre os biomarcadores mapeados, GST, GPx e dano ao DNA foram os mais relevantes para prever as condições do sedimento. Os resultados obtidos reforçam que um grupo de indicadores tem maior potencial preditivo se usados em conjunto. Uma vez detectada um tipo de resposta, as análises mais detalhadas de contaminação do sedimento podem ser realizadas. Este estudo destaca ainda que estes indicadores biológicos poderiam ser utilizados constantemente em programas de monitoramento, com a finalidade de otimizar as análises.
In this study, the association between metal concentration, toxicity in target organisms, and biomarkers was investigated. The objective was to test whether biological indicators (biomarkers, bioaccumulation and toxicity) can predict pollutant concentrations in the sediment. Five previously published studies formed the data set. For the analyses, only variables common to all studies were considered. In total, there were 9 sediment variables and 15 biological indicators. The data was analyzed using machine learning algorithms. 80% of the data was used to train the model and 20% for testing. After standardization, contamination data from all studies were grouped into five sets. The results of the correlation analyses showed the absence of correlation between the indicators, highlighting the importance of analyzing them together. In the training phase, the machine learning model selected a subset of biological indicators to predict the sediment pollution gradient, obtaining 91.479% accuracy. In the test set this accuracy rose to 100%. Toxicity occurred mainly in groups with the highest concentration of metals, suggesting a relationship between contaminants and effects on organisms. Among the mapped biomarkers, GST, GPx and DNA damage were the most relevant for predicting sediment conditions. The results obtained reinforce that a group of indicators has greater predictive potential if used together. Once a response type is detected, more detailed sediment contamination analyses can be performed. This study also highlights that these biological indicators could be used constantly in monitoring programs, to optimize analyses.
In this study, the association between metal concentration, toxicity in target organisms, and biomarkers was investigated. The objective was to test whether biological indicators (biomarkers, bioaccumulation and toxicity) can predict pollutant concentrations in the sediment. Five previously published studies formed the data set. For the analyses, only variables common to all studies were considered. In total, there were 9 sediment variables and 15 biological indicators. The data was analyzed using machine learning algorithms. 80% of the data was used to train the model and 20% for testing. After standardization, contamination data from all studies were grouped into five sets. The results of the correlation analyses showed the absence of correlation between the indicators, highlighting the importance of analyzing them together. In the training phase, the machine learning model selected a subset of biological indicators to predict the sediment pollution gradient, obtaining 91.479% accuracy. In the test set this accuracy rose to 100%. Toxicity occurred mainly in groups with the highest concentration of metals, suggesting a relationship between contaminants and effects on organisms. Among the mapped biomarkers, GST, GPx and DNA damage were the most relevant for predicting sediment conditions. The results obtained reinforce that a group of indicators has greater predictive potential if used together. Once a response type is detected, more detailed sediment contamination analyses can be performed. This study also highlights that these biological indicators could be used constantly in monitoring programs, to optimize analyses.
Descrição
Citação
FONSECA, Marina Ferrel. Associação entre indicadores ecotoxicológicos e de poluição do sedimento. 2023. 57 f. Dissertação (Mestrado Interdisciplinar em Ciência e Tecnologia do Mar) - Universidade Federal de São Paulo, Instituto do Mar, Santos, 2023.