Machine Learning and Deep Learning Methods in Ecotoxicological QSAR Modeling

What is it about?

Today the registered chemical structures are about 28 millions, while experimental toxicity data are available for a few hundred thousands of them. Defining properties and effects for all the available chemicals is such a huge task that predictive modeling is the only solution, well represented by Quantitative Structure-Activity Relationships (QSARs) models. Modeling human CMR toxicity (Carcinogenic, Mutagenic, Reproductive) and environmental PBT properties of chemicals (Persistent, Bioaccumulative and Toxic), can be done using the advancements of technology and machine learning (ML) in particular. ML has provided the tools to create systems that can think, solve problems, find patterns; these new methods are named deep learning (DL). After reviewing the state of the art, we present a DL-based QSAR model for predicting mutagenicity with the Ames test. The results obtained challenge the current state of the art. In addition, the system does not use any chemistry knowledge besides the 2D molecular structures. This opens the way to the construction of models that learn only from data and can discover patterns in data.

Featured Image

Photo by Linhao Zhang on Unsplash

Photo by Linhao Zhang on Unsplash