FRONTEO and Daiichi Sankyo Develop Predictive Model to Automatically Extract Toxicological Interpretations from Toxicity Test Reports

FRONTEO Inc. (Headquarters: Minato-ku, Tokyo; President and CEO: Masaaki Morimoto; hereinafter "FRONTEO") has successfully developed and validated a predictive model in collaboration with Daiichi Sankyo Company, Limited (hereinafter "Daiichi Sankyo") using its proprietary equation-driven AI "KIBIT"*1. This model automatically extracts descriptions related to "toxicological interpretations" from both "SEND*2 data" and "toxicity test reports."

The results were presented at the "53rd Annual Meeting of the Japanese Society of Toxicology," held from July 1 to 3, 2026, where both companies' research teams delivered a presentation titled "The Potential of Natural Language Processing in Text Analysis of Toxicity Test Reports."

Overview of Toxicity Information Analysis Initiative

This project is part of a broader initiative launched in November 2024 by FRONTEO and Daiichi Sankyo, applying the technology from FRONTEO's AI drug discovery support service "Drug Discovery AI Factory (DDAIF)"*3 to analyze and utilize toxicity test databases and reports*4,5.

To leverage insights from past toxicity tests for future R&D and safety assessments, both companies have been developing and technically validating models for a system to collect and manage toxicity test data. At last year's 52nd Annual Meeting, they reported on a foundational system that stores SEND data and toxicity test report texts in a database, enabling unified access on a single platform.

This time, as the next step, they verified the feasibility of a technology that links, via AI, the "numerical data" stored in databases with the "toxicological interpretations" described in test reports. It was confirmed that relevant descriptions could be extracted with nearly equivalent accuracy not only from standardized English reports for regulatory submissions but also from Japanese reports of exploratory toxicity tests with diverse formats, demonstrating versatility across languages and document types.

Development and Validation Results of the Predictive Model

In pharmaceutical R&D, toxicity tests are conducted to assess the safety of candidate compounds. Numerical data such as blood test results and textual data such as pathological findings are stored in databases in SEND format. Meanwhile, expert toxicological interpretations such as "this finding is drug-related" or "this change is not toxicologically significant" are recorded separately as text in test reports.

To efficiently reuse accumulated data, it is essential to automatically link these "numerical data," "textual data," and "expert interpretations." However, due to the diverse formats of reports, linking them has been challenging.

In this study, FRONTEO used "KIBIT" to develop a predictive model targeting Daiichi Sankyo’s toxicity test database and reports. The model aims to extract descriptions related to toxicological interpretations from reports to automatically link them with SEND data.

As a result, effective predictive models were successfully built for both "English reports of regulatory-compliant tests" with structured and detailed descriptions and "Japanese reports of exploratory toxicity tests" with diverse structures and concise descriptions, confirming relatively high accuracy in extracting relevant statements.

Future Outlook

Pharmaceutical companies have accumulated vast amounts of toxicity test data over many years, but much of it is stored as individual test reports, and cross-organizational reuse of insights remains insufficient. The predictive model developed in this study enables systematic organization and retrieval of past toxicity data and expert interpretations, opening up applications across multiple stages of drug discovery, such as predicting the safety of new compounds and optimizing toxicity test designs.

Currently, Daiichi Sankyo is advancing the development of a toxicity test database and a report utilization platform to leverage past toxicity test data for future R&D and safety assessments. Both companies are now exploring the development of systems to effectively utilize accumulated knowledge by implementing the results of this study on this platform.

FRONTEO will continue to contribute to the efficient development of safe medicines and the improvement of healthcare quality and patient QOL through collaboration with researchers and the provision of DDAIF.

*1 Equation-driven AI "KIBIT": A proprietary artificial intelligence developed by FRONTEO. By using equations, it enables discontinuous discoveries, causal relationship identification, and high reproducibility. Its lightweight learning process allows high-speed, high-precision analysis at the CPU level.

*2 SEND: Standard for Exchange of Nonclinical Data. A standard format for nonclinical study data established by the Clinical Data Interchange Standards Consortium (CDISC), an international nonprofit organization promoting data standardization in pharmaceutical and medical device development.

*3 DDAIF: An AI-powered drug discovery support service where FRONTEO’s drug discovery experts, proficient in AI and drug development, leverage KIBIT’s natural language processing and proprietary analytical methods to provide hypotheses for target molecule and indication discovery.

*4 Press release dated November 12, 2024: Daiichi Sankyo and FRONTEO Sign Contract for Toxicity Information Optimization and Analysis Using Drug Discovery AI Factory, https://www.fronteo.com/pr/20241112

*5 Press release dated August 18, 2025: FRONTEO and Daiichi Sankyo Sign Phase 2 Contract for Toxicity Information Analysis Using Drug Discovery AI Factory, https://www.fronteo.com/news/pr/20250818

About FRONTEO Drug Discovery AI Factory (DDAIF)

[Reference: Collaborations with Pharmaceutical Companies and Academia]

URL: https://www.fronteo.com/news/ddaif-list

"FRONTEO Drug Discovery AI Factory (DDAIF)" is an AI-powered drug discovery support service that integrates the equation-driven AI "KIBIT (Kibbit)" specialized in natural language processing (patented in Japan, Europe, the U.S., and South Korea*) with the expertise of FRONTEO’s drug discovery researchers and AI engineers. By analyzing disease-related gene networks and constructing hypotheses for candidate target molecules, DDAIF strongly supports researchers’ decision-making in drug development. The service has already been adopted and proven effective by multiple major pharmaceutical companies. *The technologies used in Drug Discovery AI Factory are protected by 21 patents granted by FRONTEO in Japan, Europe, the U.S., and South Korea.

URL: https://lifescience.fronteo.com/products/drug-discovery-ai-factory/

About FRONTEO Inc. URL: https://www.fronteo.com/

FRONTEO supports experts across various fields who face societal challenges every day by providing its in-house developed equation-driven AI "KIBIT (Kibbit)," fostering innovation