ARTEMIS enables reusable analytical context generation from scientific literature

Abstract Background Regulatory-science questions often require synthesis across large, heterogeneous, and rapidly evolving scientific literature. In domains such as monoclonal antibody development, the same body of evidence may need to support different analytical goals, including high-level question answering, technology assessment, ethical review, and historical interpretation. Modern large language models (LLMs) can assist with such tasks, but their usefulness depends heavily on the quality and structure of the context provided. This creates a need for approaches that generate reusable analytical contexts from literature rather than relying on direct access to raw documents. Methods We present ARTEMIS (Automated Research & Trend Extraction via Memory-based Insight Synthesis) in a use case focused on monoclonal antibody literature analysis. ARTEMIS transforms a chronological document corpus into reusable analytical contexts through three core capabilities: comprehensive evidence-grounded synthesis across the literature, persona-guided synthesis that supports different perspectives from the same source base, and timeline-aware context generation for historical synthesis and trend interpretation. Results In this case study, ARTEMIS generated reusable analytical contexts that supported multiple forms of literature analysis. As one downstream demonstration, ARTEMIS-based contexts provided a more effective basis for high-level question answering than raw-chunk retrieval, achieving higher answerability (100% vs. 83%) and higher or comparable quality scores in an LLM-as-a-Judge evaluation while using fewer input tokens on average. Beyond Q&A, ARTEMIS supported persona-guided synthesis by producing distinct analyses from the same corpus under three expert personas (a 3Rs Ethical Reviewer, a Clinical Development Risk Strategist, and a NAMs Technology Analyst) and enabled timeline-aware historical synthesis of topic evolution. Conclusion In this case study, ARTEMIS provided useful support for literature-analysis challenges in regulatory science that require grounded question answering, persona-guided synthesis, and historical interpretation from a shared evidence base.