Image

AnimalGAN | SafetAI | BERTox | PathologAI

Objective: To apply the large language models (LLMs) such as BERT and GPT to facilitate analysis of FDA documents and public literature for improved efficiency and accuracy of information retrieval and toxicity assessment.

Introduction: FDA has historically generated and continues to generate multiple documents during the product-review process, which contain text that cannot be readily indexed or mapped onto standard database fields and often lack metadata. Therefore, analysis of semantic relationships plays a vital role in extracting useful information from these FDA documents to facilitate regulatory science research and improve the FDA product-review process. Meanwhile, AI-based Natural Language Processing (NLP) has been focused on developing LLMs trained with large text documents to perform a broad range of NLP tasks. This initiative aims to assess the application of LLMs for FDA documents as well as developing content-specific LLMs to facilitate regulatory science at FDA such as information retrieval and text summarization.

Approaches: BERTox is a suite of NLP applications powered by diverse functions ranging from information retrieval, sentiment analysis, text classification, and Name Entity Recognition (NER). In several pilot studies, the BERTox approach has been applied to drug-induced liver injury classification based on FDA drug labeling, causal inference of the FDA Adverse Event Reporting Systems (FAERS) database, AI bias in interpretation and classification of drug properties (e.g., safety and efficacy), text summarization to provide highlights of labeling sections, and automatic anomaly analysis. The initiative has a specific emphasis on developing responsible AI models with customized LLMs that can be operated in a local environment for specific regulatory applications with understanding of their bias, context of use, causal inference, and explainability.

Potential impact: Reviewing text documents is a crucial step in assessing the safety and efficacy of FDA-regulated products. However, the current manual process is time consuming and resource intensive. BERTox offers a set of LLMs-based AI tools/systems to intelligently process and extract critical information from FDA documents to improve and expedite the product-review process. In addition, BERTox can also serve as an institutional memory to effectively access past documents that are often referenced to ensure consistency and evidence-based decision-making in the review of new products.

Resources for You

• Artificial Intelligence at NCTR
• NCTR Research Focus Areas
• Bioinformatics Tools
• Division of Bioinformatics and Biostatistics