• The importance of scaling in data mining for toxicity prediction.

      Mazzatorta, P.; Benfenati, E.; Neagu, Daniel; Gini, G. (2002)
      While mining a data set of 554 chemicals in order to extract information on their toxicity value, we faced the problem of scaling all the data. There are numerous different approaches to this procedure, and in most cases the choice greatly influences the results. The aim of this paper is 2-fold. First, we propose a universal scaling procedure for acute toxicity in fish according to the Directive 92/32/EEC. Second, we look at how expert preprocessing of the data effects the performance of qualitative structure-activity relationship (QSAR) approach to toxicity prediction.
    • Ranking strategies to support toxicity prediction: a case study on potential lxr binders

      Palczewska, Anna Maria; Kovarich, S.; Ciacci, A.; Fioravanzo, E.; Bassan, A.; Neagu, Daniel (2019-05)
      The current paradigm of toxicity testing is set within a framework of Mode-of-Action (MoA)/Adverse Outcome Pathway (AOP) investigations, where novel methodologies alternative to animal testing play a crucial role, and allow to consider causal links between molecular initiating events (MIEs), further key events and an adverse outcome. In silico (computational) models are developed to support toxicity assessment within the MoA/AOP framework. This paper focuses on the evaluation of potential binding to the Liver X Receptor (LXR), as this has been identified among the MIEs leading to liver steatosis within an AOP framework addressing repeated dose and target-organ toxicity. The objective of this study was the development of a priority setting strategy, by means of in silico approaches and chemometric tools, to allow for the screening and ranking of chemicals according to their toxicity potential. As a case study, the present paper outlines the methodologies and procedures that have been developed in the context of the COSMOS/cosmetics safety assessment project [4], which developed computational methods in view of supporting cosmetics safety assessment, to rank chemicals based on their potential binding to LXR. Chemicals are ranked based on molecular and QSAR modelling outcomes. The contribution in this paper is threefold: the QSAR model for LXR dataset, an application of molecular modeling approaches, which have been developed and optimized for drug discovery, in the context of toxicology, and finally ranking chemicals based on diverse modelling outcomes. The novelty in this paper consists of the employment of linear (logistic regression) and non-linear (Random Forest) models in the context of ranking chemicals. The results show that these methods can be successfully applied for prioritization of compounds of major concern for potential liver toxicity, and that they perform better than the ranking methods reported in the literature to date (such as total ordering or data fusion).