PBT assessment and prioritization by PBT Index and consensus modeling: comparison of screening results from structural models.

The limited availability of comprehensive data for Persistence, Bioaccumulation and Toxicity (PBT) of chemicals is a serious hindrance to the assignment of compounds to the categories of PBT and vPvB; REACH regulation requires authorization for the use of such chemicals, and additionally plans for safer alternatives. In the context of screening and priority-setting tools for PBT-assessment, the cumulative PBT Index model, implemented in QSARINS (QSAR-INSUBRIA), new software tool for the development and validation of multiple linear regression QSAR models, offers a new holistic approach for the identification of chemicals with cumulative PBT properties directly from their molecular structure. In this study the Insubria PBT Index in QSARINS is applied to the screening and prioritization of various data sets, containing a large variety of chemicals of heterogeneous molecular structure, previously screened by various authors by different methods, for their potential PBT behavior. Particular attention is devoted to the model Applicability Domain, using different approaches such as Descriptor Range, Leverage, and Principal Component Analysis (PCA) of the modeling molecular descriptors, in order to discriminate between interpolated and extrapolated predictions. The results of this screening, which is based only on the molecular structure features and is not dependent on single threshold values for P, B and T, are compared with those obtained by the on-line US-EPA PBT Profiler. Good agreement between the various approaches is found, supporting the utility of a consensus approach in priority-setting studies. The main discrepancies are highlighted and commented on. Moreover, a priority list containing the most hazardous compounds identified in agreement between the two tools is drafted. The PBT Index, implemented in QSARINS, which was demonstrated to be a practical, precautionary and reliable screening tool for PBT-behavior directly from the molecular structure, can be usefully applied for focusing experimental studies, and even before chemical synthesis, in a "benign by design" approach of safer alternatives.