Affordable Access

Comprehensive data analysis and predictive chemoinformatics models for REACH related physicochemical and (eco)toxicity properties

Authors
  • Lunghini, Filippo
Publication Date
Sep 29, 2020
Source
HAL
Keywords
Language
English
License
Unknown
External links

Abstract

This thesis concerns the modelling of several environmental fate and (eco)toxicological properties relevant under the European Union Registration, Evaluation, Authorisation and Restriction of Chemical Substances Regulation (REACH, EC No 1907/2006). Statistical models have been generated using state-of-the-art machine learning methods, such Support Vector Machine and Random Forest and molecular descriptors. Models have been internally and externally validated following internationally recognized guidelines, especially the OECD principles. The models are designed to be used as valid alternative to experimental testing and data-gap filling under the REACH regulation. New models possess several advantages over already existing ones: (i) noticeable larger training sets; (ii) external validation on a significant number of compounds coming from the Industrial context (Solvay portfolio); (iii) better accuracy and extended applicability domain.

Report this publication

Statistics

Seen <100 times