ABSTRACTREACH requires that all chemicals produced or imported for more than 1 ton/year must be tested for ready biodegradability, a screening test for the assessment of chemical persistence. Since REACH encourages the use of in-silico approaches, several Quantitative Structure-Activity Relationship models have been calibrated to predict this property. Predictions obtained by different models can then be merged by means of specific consensus approaches, because it has been demonstrated that data integration can increase reliability and reduce the effects of contradictory data by averaging results.This study deals with integration of ready biodegradability predictions from eight different models. A set of 416 molecules, external to the training sets of all the considered models, was retrieved from the literature. Predictions were combined with three consensus approaches: majority voting criterion, Dempster–Shafer's theory of evidence and Bayesian consensus with discrete probability distributions. The different approaches were finally evaluated on the basis of the classification performance.On the basis of the obtained results we can conclude that application of consensus methodologies resulted in a reduction of uncertainty associated with biodegradability predictions and showed apparent advantages when sources of information are jointly analysed, without compromising the quality of predictions.