AI as a game changer in toxicology?

Artificial intelligence (AI) is probably familiar to everyone by now. It has already found its way into many areas, such as finance, education, medicine and many other fields. Due to its ability to process and analyse large and complex amounts of data, it would also be well suited for use in the life sciences. For example, the use of AI would facilitate and perhaps even improve the development of medical issues, from diagnosis to therapy, the implementation of studies in drug discovery, toxicology or genetics.

The scientific discipline of toxicology is concerned with observing, analysing and describing the effects of a certain agent in a living organism or ecosystem. It therefore plays an important role in testing the safety of new products, e.g. in the development of medicines, but also in the prevention of diseases. While AI systems have long been used in other scientific fields, there is still some catching up to do in toxicology. For a long time, most toxicological studies were based on results from in-vivo experiments. However, as toxicology has now become considerably richer in data compared to the past, the use of AI could certainly be considered useful. Artificial intelligence could be used to predict toxic reactions, which could be used to assess potential risks. It would also be much easier and quicker to analyse large amounts of data.

The integration of AI was much faster in pathology. High-resolution imaging methods for more precise examination of tissue samples or various digital tools for automated quantitative and objective analysis of cell count, morphology and colour intensity of cells, for example, would be examples of this. Digital pathology could be combined well with toxicology. For example, it would be suitable for screening the effects of various substances in tissue. The advantages would be that data could be shared more easily because digital histological preparations could be accessed from anywhere. Furthermore, automated analyses would lead to considerable savings in time and money. Digital pathology makes a significant contribution to the Reduction and Replacement of animal experiments because it can be easily integrated into in-vitro and in-silico models. In the case of animal experiments, which cannot yet be replaced by alternative methods, the high precision of digital pathology could lead to better and more accurate results.

Apart from the many benefits that the integration of AI brings, there are still a few limitations that need to be worked on in the future. Not all available data should be included in data analyses. Falsified, inaccurate and otherwise unusable sources would only distort the results of the analyses. Mechanisms would therefore still be needed to sort out the "rubbish". AI systems should also deliver comprehensible and transparent results. Current systems can be compared to a "black box", which makes interpretation considerably more difficult. In order to ensure the reproducibility of tests, standardised AI models should also be developed.

Source: Hartung, Thomas. (2023). ToxAIcology - The evolving role of artificial intelligence in advancing toxicology and modernizing regulatory science. ALTEX. 40. 559-570. 10.14573/altex.2309191.