Artificial intelligence and Artificial Neural Networks in toxicology: challenges, perspectives and applications (Narrative review)

Abstract

Artificial Neural Networks (ANN) have great potential in toxicology research. They may be used to predict the toxicity of various chemical compounds or to classify compounds based on the toxic effects they have on the body or in the environment. Nowadays, numerous ANN models have been developed, some of which may be used to identify and possibly explain complex chemical-biological interactions in toxicological sciences. Tightly coupled multilayer perceptron’s may, under some conditions, have high classification accuracy and discrimination power in separating damaged from intact cells after exposure to a toxicant. Regularized and fully connected convolutional neural networks cannot detect and detect discrete changes in toxicity related two-dimensional data patterns. Bayesian neural networks with weight marginalization may sometimes have better prediction performance compared to traditional approaches. With the further development of artificial intelligence, artificial neural networks are expected to become an important part of various accurate and cost-effective biosensors for detecting various toxic substances and evaluating their biochemical properties in the future. In this brief review article, we discussed the recent researches focused on the scientific value of ANN in evaluating and predicting the toxicity of various chemical compounds and the challenges and limitations in the field of artificial intelligence and the use of artificial neural networks.

Keywords

• artificial intelligence
• artificial neural networks
• toxicology