Different KNIME workflows for read-across and successive use for weight-of-evidence strategy
Résumé
The evaluation of the toxic effects of substances is a complex task, due to the huge amount of factors involved in the biological processes at the basis of the effect. This requires taking advantage of all elements that can be used in the assessment of the property values. The read-across approach and the in silico methods, collectively called non-testing methods, can be integrated within a weight-ofevidence strategy. This integration is typically performed manually. Furthermore, also the read-across process in most of the cases relies on expert decisions, which may be subjective, and based on some initial choices. In this approach, there is a risk of making poorly reproducible results besides losing important pieces of information. In addition, a main shortcoming in read-across is that the process may not identify some of the relevant source compounds. In order to cope with these problems, we explored software tools able to assist the expert. The factors related to similarity which we used to select source compounds were: structural, physico-chemical, toxicological and pharmacokinetic features. These tools analyse the similarities of the compounds in “full or partial” way, i.e. merging all the features or selecting only those more relevant. Furthermore, the steps of the process can be done in a parallel or sequential way. Finally, we combined the results of the read-across procedure with those from in silico models. We will describe the added value of these programs, implemented in KNIME. We acknowledge the project EU-ToxRisk (a project funded by the European Union’s Horizon 2020 research and innovation program under grant agreement No 681002).