Probabilistic graphical models for the identification and analysis of reaction pathways in water treatment
Résumé
In this paper, we proposed a new pathway enrichment analysis for reaction schemes based on Bayesian Network (BN). Several reaction schemes are proposed in the waster waters treatment literature, however, these data diverge greatly and, to date, are unexploited. As an innovative alternative, probabilistic graphical models could allow establishing parentage between the most relevant molecules and Transformation Products (TPs). Based on the analysis of an extensive bibliography gathering more than 45 articles for more than 140 molecules, this study proposed an innovative methodology based on knowledge graphs, pathway data enrichment, and Bayesian Networks (BN) analyses. The proposed methodology has been applied to elucidate the degradation of the sulfamethoxazole. Probabilistic approaches and graphical models bring a new light on the identification of the TP parentage.