EVENT DOSSIER
On September 7, 2026, the Hakken system was officially released. Its purpose is to fill current knowledge gaps through predicting future scientific discoveries. The system constructs prediction models based on time-series knowledge graphs and semantic knowledge from large language models, and utilizes an explanation framework to assist scientists in evaluating new relationships. The research team applied it to the field of biomedicine, established a benchmark for predicting time-sensitive multi-label relationships, and verified the model’s consistency in long-time-series data. Additionally, the team identified 1.5 million highly confident hypotheses. After qualitative verification by biologists, three wet experiments were conducted, and two new interactions with significant impact on drug discovery were confirmed: the new interactions between TP53 and BAMBI, and RAF1 and TNF.
Hakken 系统发布，旨在通过预测未来发现填补当前知识空白。该系统基于时间序列的知识图谱与大型语言模型语义知识构建预测模型，并调用解释框架辅助科学家评估新关系。研究团队将其应用于生物医学领域，建立了时间感知多标签关系预测基准，验证了模型在长时序数据中的连贯性。此外，团队筛选出 150 万个高置信度假设，经生物学家定性验证后推进三项湿实验，最终确认两项对药物发现具有重大影响的预测：TP53 与 BAMBI、RAF1 与 TNF 之间的新相互作用。