arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2601.20906cs.LG
收藏

TwinWeaver: 一种基于大语言模型的跨癌症数字双胞胎基础模型框架

TwinWeaver: An LLM-Based Foundation Model Framework for Pan-Cancer Digital Twins

Computational Sciences Center of Excellence, Roche, Penzberg, Germany(罗氏计算科学卓越中心);Computational Health Center, Helmholtz Munich, Munich, Germany(海德堡慕尼黑计算健康中心);Department of Biology, Ludwig Maximilian University of Munich, Munich, Germany(慕尼黑路易斯·马克西米利安大学生物学系);Early Development Oncology, Roche Innovation Center Zurich, Roche, Schlieren, Switzerland(罗氏苏黎世创新中心早期肿瘤学);Computational Sciences Center of Excellence, Genentech, New York City, USA(基因泰克计算科学卓越中心);Computational Sciences Center of Excellence, Genentech, South San Francisco, USA(基因泰克计算科学卓越中心);Center for Data Science, New York University, New York City, USA(纽约大学数据科学中心);Department of Computer Science, Stanford University, Stanford, CA, USA(斯坦福大学计算机科学系);Computational Sciences Center of Excellence, Roche, Basel, Switzerland(罗氏巴塞尔计算科学卓越中心);Department of Biochemistry(生物化学系);Biotechnology Institute, The University of Melbourne, Melbourne, Australia(墨尔本大学生物技术研究所)
Nikita Makarov, Maria Bordukova, Lena Voith von Voithenberg, Estrella Pivel-Villanueva, Sabrina Mielke, Jonathan Wickes, Hanchen Wang, Mingyu Derek Ma, Keunwoo … 展开作者

Nikita Makarov, Maria Bordukova, Lena Voith von Voithenberg, Estrella Pivel-Villanueva, Sabrina Mielke, Jonathan Wickes, Hanchen Wang, Mingyu Derek Ma, Keunwoo Choi, Kyunghyun Cho, Stephen Ra, Raul Rodriguez-Esteban, Fabian Schmich, Michael Menden

AI总结:

TwinWeaver通过基于大语言模型的框架,构建跨癌症数字双胞胎,提升临床事件预测与风险分层精度。

摘要:

精准肿瘤学需要预测临床事件和轨迹,但建模稀疏的多模态临床时间序列仍然是一个关键挑战。我们介绍了TwinWeaver,一个开源框架,将纵向患者病史序列化为文本,使能够统一预测事件以及利用大语言模型进行预测,并用它在20种癌症类型上的93054名患者上构建Genie数字双胞胎(GDT)。在基准测试中,GDT显著降低了预测误差,达到中位均方误差(MASE)为0.87,相比最强的时间序列基线(MASE为0.97)(p<0.001)。此外,GDT提高了风险分层,平均一致性指数(C指数)为0.703,覆盖生存、进展和治疗切换任务,超过最佳基线0.662。GDT还泛化到分布外的临床试验,零样本匹配训练基线,微调后超过基线,达到中位MASE为0.75-0.88,并在事件预测中以平均C指数0.672优于最佳基线0.648。最后,TwinWeaver使临床推理可解释性扩展成为可能,提供了一种可扩展且透明的基础,用于纵向临床建模。

英文摘要:

Precision oncology requires forecasting clinical events and trajectories, yet modeling sparse, multi-modal clinical time series remains a critical challenge. We introduce TwinWeaver, an open-source framework that serializes longitudinal patient histories into text, enabling unified event prediction as well as forecasting with large language models, and use it to build Genie Digital Twin (GDT) on 93,054 patients across 20 cancer types. In benchmarks, GDT significantly reduces forecasting error, achieving a median Mean Absolute Scaled Error (MASE) of 0.87 compared to 0.97 for the strongest time-series baseline (p<0.001). Furthermore, GDT improves risk stratification, achieving an average concordance index (C-index) of 0.703 across survival, progression, and therapy switching tasks, surpassing the best baseline of 0.662. GDT also generalizes to out-of-distribution clinical trials, matching trained baselines at zero-shot and surpassing them with fine-tuning, achieving a median MASE of 0.75-0.88 and outperforming the strongest baseline in event prediction with an average C-index of 0.672 versus 0.648. Finally, TwinWeaver enables an interpretable clinical reasoning extension, providing a scalable and transparent foundation for longitudinal clinical modeling.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑