SK Biopharm Builds AI Model Specialized in Drug Development, Plans Real-World Research Applications

Participated in consortium with Lunit to develop and validate a model specialized in drug development
Leveraging AI infrastructure powered by NVIDIA B200 GPUs

SK Biopharm said on Nov. 6 that the “AI-Specialized Foundation Model Project,” led by the Ministry of Science and ICT and involving the company, had completed about 10 months of development and demonstration.

SK Biopharm participated in every stage of the project, from building datasets and developing AI models and agents to demonstrating their use in new drug development and fostering an ecosystem. A consortium comprising companies including SK Biopharm, Lunit, Trillion Labs, Aigen Sciences and Kakao Healthcare, along with domestic academic, industry and medical institutions, built a foundation model specialized in medical science. Based on this model, the consortium developed specialized models in three areas: ▲drug development (compounds and proteins) ▲medical science (omics) ▲clinical applications.

SK Biopharm corporate logo image. SK Biopharm

From the early stages of the project, SK Biopharm participated in building biological datasets covering small-molecule compounds, proteins and genomes, as well as developing and validating AI models specialized in new drug development. Drawing on its expertise in pharmaceutical research and development (R&D) and the needs of researchers in the field, the company built specialized training data for new drug development totaling approximately 7.14 million records and 4.4 billion tokens. The data covered 66 new drug development tasks, including drug properties and toxicity, drug-target interactions, drug responsiveness and clinical trial outcome prediction, and was used to train the models and validate their performance.

SK Biopharm also validated the performance of a model specialized in new drug development that it co-developed with Trillion Labs. Using 66 tasks as benchmarks, it evaluated predictive performance across different areas of new drug development, including drug properties, toxicity, drug-target interactions, drug responsiveness and clinical trial outcome prediction. The company also conducted additional validation using confidential drug data from actual pharmaceutical companies to determine whether the models could use not only public data but also data accumulated during real-world new drug development.

The company also explored the potential of AI in pharmaceutical R&D and clinical development. It built an AI work-support system that assists with tasks ranging from extracting information on research chemicals and expert review to generating submission documents. It also developed a digital twin model using publicly available Alzheimer's disease clinical data to predict changes in individual patients' cognitive function over time and update the predictions when new visit information is entered. These efforts demonstrated the potential for using AI to analyze patient prognoses and select and design clinical trial participants.

Building on the capabilities it gained in developing specialized foundation models through the project and its recently established AI computing infrastructure powered by NVIDIA B200 GPUs, SK Biopharm plans to step up its “TR-AX” initiative. The company plans to apply specialized foundation models to actual research projects to strengthen its capabilities in researching diseases, targets and compounds, and to expand the use of AI from identifying drug candidates through clinical development, boosting R&D productivity and execution.

“The competitiveness of AI for new drug development ultimately depends on how meaningful the answers it provides are in actual research settings,” said SK Biopharm President Lee Donghoon. “Through this project, we directly integrated pharmaceutical companies' data and research expertise into AI models and confirmed their potential in real-world new drug development scenarios.” He added, “Going forward, we will step up TR-AX by building on our in-house AI infrastructure and specialized model capabilities, enabling AI to meaningfully enhance researchers' decision-making and develop into a core competitive advantage that increases the speed and likelihood of success in new drug development.”

This content was produced with the assistance of AI translation services.