Application of Bayesian Network modeling on the stability and toxicity of engineered nanomaterials in aquatic ecosystems

Abstract

The stability of engineered nanomaterials (ENMs) in the aquatic systems influences their eventual interactions with aquatic biota – and subsequently the observed toxic effects. Increasing data suggests that physicochemical properties of ENMs such as size, surface charge, chemical composition etc. can be enhanced, modified, or neutralized owing to biotic and water chemistry factors such as organic and inorganic substrates, and ionic strength. When ENMs enter aquatic systems, they undergo transformation, and acquire dynamic properties such as dispersion, agglomeration/aggregation, and sedimentation which control their interactions with organisms in the water column and sediments. In this study, stability and toxicity data of ENMs solicited from published scientific reports was used to develop a Bayesian Network (BN) model to predict their potential risks to the aquatic organisms. The suitability of the BN modeling tool was exploited to investigate the different plausible scenarios of complex interactions of multiple causal and effect variables. Likely states of each variable had a predictive evidence of = 1 probability. The model was designed in a modular version where each sub-model evaluated a single dynamic property of ENMs in a stepwise manner. Here, the functionality of the developed model is illustrated using data for different types (e.g. anatase, rutile) and forms (bare, coated, pristine, and formulated) of nTiO2. Both the prior and posterior models of the BN were developed using collected and collated data obtained from scientific published literature. Our findings demonstrate predictive reasoning capability to link causes and effects of ENMs in the aquatic systems. Depending on the degree of influence of interacting variables, high probability outcomes from our results indicate high likelihood of the causal variables influencing the succeeding effects. The BN tool applied here illustrates its suitability to evaluate and predict the potential the risks of ENMs in aquatic systems using probability methods.