The government has developed a technology that utilizes artificial intelligence (AI) to automatically find extensive toxicity data on chemicals and create a database. The aim is to automate the entire chemical risk assessment process by 2030, reducing the evaluation time by over 70%.
The National Institute of Environmental Research, under the Ministry of Climate, Energy, and Environment, announced on October 7 that it has developed a technology for AI to automatically collect and extract toxicity data used in chemical risk assessments and will file for an international patent.
Risk assessment is the process of evaluating the harmfulness of chemicals and their impact on humans and the environment. It involves several steps: hazard identification, dose-response assessment, exposure assessment, and risk characterization.
Previously, experts had to manually search through extensive domestic and international reports and documents to review and transcribe necessary toxicity values, limiting the number of substances that could be evaluated. The new technology automates three stages of the hazard identification process, from report investigation and collection to toxicity data extraction and database creation.
The AI automatically accesses institutional websites to gather report lists and detailed information, saving the original documents. It then analyzes the table of contents of lengthy reports to identify sections related to human and ecological toxicity, automatically extracting information such as test species and toxicity values.
Links to the original documents are also generated for the extracted data, allowing evaluators to immediately compare the AI's results with the original materials, thereby verifying the reliability of the findings.
The National Institute of Environmental Research has developed a pilot model called 'ATOM (AI-based Toxicological evidence Miner)' to collect and extract data from risk assessment reports in the European Union (EU).
In the future, the scope of automation will gradually expand beyond hazard information collection to include exposure assessment, risk characterization, and report writing. The goal is to establish a fully automated risk assessment system by 2030, reducing the evaluation period for each substance by more than 70%.
Currently, the National Institute evaluates about 15 chemical substances annually. If the institute's goal of tripling this number is applied, it would mean that at least 45 substances could be evaluated annually after automation.
The developed technology has already been registered as a domestic patent under the title 'Method and System for Fully Automated Collection, Extraction, Synthesis, and Database Creation of Toxicity Data Based on Artificial Intelligence.' The National Institute is also pursuing an international patent based on this.
The government plans to expand related technological cooperation through the International Coalition for Accelerating Risk Assessment (APCRA), which includes governments from 10 countries, as well as the United Nations (UN), the Organisation for Economic Co-operation and Development (OECD), and the EU.
Kim Soo-jin, head of the Environmental Health Research Department at the National Institute of Environmental Research, stated, "With this international patent application, we have laid the foundation for quickly and consistently deriving information from vast toxicity data. We believe we can expand the number of evaluated substances from the current approximately 15 to more than three times that in the future."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.