❌

Normal view

Received — 5 March 2026 ⏭ Favourites

Exploration of multi-omics liquid biopsy approaches for multi-cancer early detection: The PROMISE study

✇(Multiomics OR Omics) AND (Pancreatic)
By: Jianchun Duan · Qiang Gao · Zhijie Wang · Jiachen Xu · Yuzi Zhang · Yanan Wang · Xi Yang · Lei Zhang · Yu Xu · Qiaoxia Zhou · Bo Yang · Xingyu Yang · Guoqiang Wang · Jing Zhao · Xuefei Wang · Di Ge · Lijie Tan · Wenju Chang · Jianmin Xu · Pinghong Zhou · Mengjiang He · Yuanlin Song · Jie Hu · Wenhui Lou · Wenchuan Wu · Tingyan Shi · Rongyu Zang · Yuan Ji · Sunfang Jiang · Youpei Lin · Liangqing Dong · Shangli Cai · Jia Fan · Jie Wang
25 February 2026 at 19:00

Innovation (Camb). 2025 Aug 6;7(1):101076. doi: 10.1016/j.xinn.2025.101076. eCollection 2026 Jan 5.

ABSTRACT

Although circulating cell-free DNA (cfDNA) methylation has emerged as the mainstream approach in multi-cancer detection blood tests (MCDBTs), the potential of integrating proteins and mutations, to enhance its performance remains unclear. The PROMISE study (NCT04972201) was conducted to investigate the feasibility of a multi-omics integration strategy in MCDBTs across nine types of cancers in head and neck (excluding nasopharynx), esophagus, lung, stomach, liver, biliary tract, pancreas, colorectum, and ovary. Blood samples were prospectively collected from 1,706 participants (840 non-cancer; 866 cancer) and then randomly divided into training and validation sets. The complementarity between various omics were investigated, and specific omics features were carefully selected for further multimodal model construction. The methylation-based classifier outperformed both the mutation-based and protein-based classifiers. As 95.0% of cancer cases detected by the mutation-based classifier were simultaneously identified by the methylation-based classifier, while 14.0% of the protein-positive samples were missed, protein markers may provide complementary value to the methylation-based classifier. Compared with the methylation-based classifier, the multimodal classifier combining methylation and protein features exhibited an improved sensitivity of 75.1% (95% confidence interval [CI], 69.3%-80.3%) at the same specificity of 98.8% with the accuracy of top predicted origin (TPO1) of 73.1% (95% CI, 66.2%-79.2%). Notably, the TPO1 accuracy reached 100% in liver and ovarian cancers with negative results of the methylation-based classifier. Collectively, these data suggest that the integration of protein markers in the multimodal classifier can offer additional benefits to the methylation-based classifier, particularly in identifying liver and ovarian cancers.

PMID:41737326 | PMC:PMC12925926 | DOI:10.1016/j.xinn.2025.101076

Received — 29 October 2025 ⏭ Favourites

BLM$_1$: A Boundless Large Model for Cross-Space, Cross-Task, and Cross-Embodiment Learning

✇cs.AI, q-bio.NC updates on arXiv.org
By: Wentao Tan · Bowen Wang · Heng Zhi · Chenyu Liu · Zhe Li · Jian Liu · Zengrong Lin · Yukun Dai · Yipeng Chen · Wenjie Yang · Enci Xie · Hao Xue · Baixu Ji · Chen Xu · Zhibin Wang · Tianshi Wang · Lei Zhu · Heng Tao Shen
29 October 2025 at 12:00
arXiv:2510.24161v1 Announce Type: new Abstract: Multimodal large language models (MLLMs) have advanced vision-language reasoning and are increasingly deployed in embodied agents. However, significant limitations remain: MLLMs generalize poorly across digital-physical spaces and embodiments; vision-language-action models (VLAs) produce low-level actions yet lack robust high-level embodied reasoning; and most embodied large language models (ELLMs) are constrained to digital-space with poor generalization to the physical world. Thus, unified models that operate seamlessly across digital and physical spaces while generalizing across embodiments and tasks remain absent. We introduce the \textbf{Boundless Large Model (BLM$_1$)}, a multimodal spatial foundation model that preserves instruction following and reasoning, incorporates embodied knowledge, and supports robust cross-embodiment control. BLM$_1$ integrates three key capabilities -- \textit{cross-space transfer, cross-task learning, and cross-embodiment generalization} -- via a two-stage training paradigm. Stage I injects embodied knowledge into the MLLM through curated digital corpora while maintaining language competence. Stage II trains a policy module through an intent-bridging interface that extracts high-level semantics from the MLLM to guide control, without fine-tuning the MLLM backbone. This process is supported by a self-collected cross-embodiment demonstration suite spanning four robot embodiments and six progressively challenging tasks. Evaluations across digital and physical benchmarks show that a single BLM$_1$ instance outperforms four model families -- MLLMs, ELLMs, VLAs, and GMLMs -- achieving $\sim\!\textbf{6%}$ gains in digital tasks and $\sim\!\textbf{3%}$ in physical tasks.
There are no more articles
❌