msc-aging-agent-corpus
收藏资源简介:
MSC衰老智能体语料库是一个经过筛选、可供智能体使用的间充质干细胞(MSC)衰老转录组关联数据集。它整合了来自公共NCBI GEO数据库的转录组数据、DVC Stem的临床研究文献证据以及BioGPS的生物数据集扩展发现信息。该数据集旨在为MSC衰老机制研究、衰老相关生物标志物发现以及跨证据层(转录组、临床、图谱)的关联分析提供结构化资源。数据内容包含多个层次:1) 核心转录组数据:涵盖19个独立的GEO数据集,提供了18,045个经过统计筛选的基因-衰老关联,每个数据集包含样本元数据、基因表达值(长格式和宽格式)以及关联分析结果(如效应值beta、p值、q值)。2) 临床证据层:包含来自DVC Stem研究数据库的54项MSC治疗临床研究的书目元数据,以及13个连接GEO转录组发现与临床文献的“桥梁”条目。3) BioGPS证据层:收录了85个BioGPS数据集的目录,并识别出98个可能与MSC/衰老研究相关的GEO扩展候选数据集,通过4个“桥梁”连接核心语料库与BioGPS图谱。数据规模介于1万到10万条之间,主要文件格式为CSV、压缩CSV和JSON。关键数据文件包括:数据集清单(dataset_manifest.csv)、临床证据清单(dvc_stem_study_manifest.csv)、转录组-临床桥梁文件(transcriptome_clinical_bridges.csv)、BioGPS目录(biogps_geo_catalog.csv)、衰老生物标志物候选基因表(aging_biomarker_candidates.csv)以及分析就绪的表达长格式表(analysis_ready_expression_long.csv.gz)。每个GEO数据集文件夹内还包含详细的样本级数据和解读说明。使用本数据集时需特别注意:必须首先加载并固定语料库的修订版本(通过CORPUS_PROVENANCE.json);需仔细阅读各数据集的解读注意事项(interpretation_caution),因为不同数据集的“衰老轴”(aging_axis)定义各异(例如,可能是供体年龄、细胞传代、发育阶段或疾病表型对比);严禁将临床摘要文本或BioGPS元数据直接等同于转录组统计证据进行合并分析。数据集适用于生物信息学分析、计算生物学建模以及需要结合多源证据的生物医学研究智能体(agent)开发。
MSC Aging Agent Corpus is a curated, agent-ready transcriptome-associated dataset focused on mesenchymal stem cell (MSC) senescence. It integrates transcriptomic data from the public NCBI GEO database, clinical study literature evidence from DVC Stem, and extended discovery information from BioGPS biological datasets. This dataset aims to provide a structured resource for research on MSC senescence mechanisms, discovery of senescence-related biomarkers, and cross-evidence-layer (transcriptomic, clinical, atlas) association analyses. The dataset content covers three layers: 1) Core transcriptomic data: It includes 19 independent GEO datasets, providing 18,045 statistically filtered gene-senescence associations. Each dataset contains sample metadata, gene expression values in both long and wide formats, and association analysis results such as effect size beta, p-value, and q-value. 2) Clinical evidence layer: It comprises bibliographic metadata for 54 MSC therapy clinical studies from the DVC Stem research database, plus 13 "bridge" entries that link GEO transcriptomic findings to clinical literature. 3) BioGPS evidence layer: It compiles a catalog of 85 BioGPS datasets, identifies 98 potential GEO extended candidate datasets relevant to MSC/senescence research, and connects the core corpus to the BioGPS atlas via 4 "bridge" entries. The dataset has a scale ranging from 10,000 to 100,000 entries, with primary file formats including CSV, compressed CSV, and JSON. Key data files are: dataset manifest (dataset_manifest.csv), clinical evidence manifest (dvc_stem_study_manifest.csv), transcriptome-clinical bridge file (transcriptome_clinical_bridges.csv), BioGPS catalog (biogps_geo_catalog.csv), candidate aging biomarker gene table (aging_biomarker_candidates.csv), and analysis-ready long-format expression table (analysis_ready_expression_long.csv.gz). Each GEO dataset folder also contains detailed sample-level data and interpretation notes. Special precautions for using this dataset are as follows: First, you must load and fix the corpus revision version via CORPUS_PROVENANCE.json; second, carefully read the interpretation cautions for each dataset, as the definition of "aging axis" varies across datasets (e.g., it may refer to donor age, cell passage, developmental stage, or disease phenotype comparison); third, it is strictly forbidden to directly combine clinical summary texts or BioGPS metadata with transcriptomic statistical evidence for combined analysis. This dataset is applicable to bioinformatics analysis, computational biology modeling, and the development of biomedical AI agents that require integration of multi-source evidence.
数据集概述:MSC Aging Agent Corpus
MSC Aging Agent Corpus 是一个经过人工筛选、面向智能体(agent)的间充质干细胞(MSC)衰老转录组关联数据集,整合了来自公共GEO数据库的转录组数据、临床证据(DVC Stem)以及BioGPS扩展发现。
数据集全称: MSC Aging Agent Corpus
发布方: Syndicate Laboratories
策展人: Dr. James Utley, PhD(长寿科学家)
核心研究焦点: 间充质干细胞(MSC)生物学作为衰老生物标志物
语言: 英语
许可协议: CC-BY-4.0
数据规模: 10,000 < n < 100,000
数据集组成
| 组成部分 | 数量 |
|---|---|
| GEO转录组数据集 | 23 |
| 筛选后的关联(保留) | 22,045 |
| 临床证据研究(DVC子集) | 54 |
| GEO ↔ 文献临床桥梁 | 15 |
| BioGPS数据集目录 | 85 |
| GEO扩展候选(MSC/衰老) | 106 |
| BioGPS桥梁(GEO ↔ 图谱) | 4 |
| 机制关联路径 | 12 |
| 机制标记的分泌组命中 | 251 |
数据来源
- 公共转录组队列: NCBI GEO
- 临床书目元数据: DVC Stem(https://www.dvcstem.com/study-database)
- 扩展发现: BioGPS(http://biogps.org/api/biogps_dataset)
核心数据文件(推荐加载顺序)
datasets/CORPUS_PROVENANCE.json— 修订版本固定、层级映射、引用模板(必须最先加载)datasets/dataset_manifest.csv— GEO来源信息、模型、解释指南datasets/clinical_evidence/dvc_stem_study_manifest.csv— 临床MSC治疗书目元数据datasets/clinical_evidence/transcriptome_clinical_bridges.csv— GEO ↔ 临床桥梁(跨层级声明前加载)datasets/biogps_evidence/biogps_geo_catalog.csv— BioGPS关联的GEO目录与扩展候选datasets/mechanistic_associations/cell_interaction_pathways.csv— MSC ↔ 伙伴细胞关系路径datasets/aging_biomarker_candidates.csv— 跨数据集基因搜索(仅表达证据)datasets/analysis_ready_expression_long.csv.gz— 用于重复运行的表达+元数据datasets/<dataset_id>/— 各GEO队列的表格与README
关键字段说明
- Manifest(清单):
dataset_id,species,cell_context,aging_axis,aging_axis_type,model_formula,expression_scale,interpretation_caution,n_samples,retained_screened_features - Biomarker table(生物标志物表):
gene_symbol,beta,p_value,q_value,direction,nominal_significant,fdr_significant+ 来源列 - Expression long(长格式表达表):
feature_id,sample_id,expression_value+ 关联统计量 + 数据集特定元数据
解释注意事项
beta对应每个数据集model_formula中的第一项(即衰老轴)。- 部分数据集存在特殊对比(如年龄分组、复制性衰老、发育轴、共培养等),需在跨数据集综合前仔细阅读对应
aging_axis_type、corpus_uniqueness_note和interpretation_caution。 - 具体数据集解释警告包括:
GSE115068(小鼠脂肪MSC周期对比)、GSE94736(老年女性组对比)、GSE34929(双色P12 vs P4)等。
引用说明
- 必须固定修订版本标签(例如
corpus-v2026.06.6) - APA引用格式: Utley, J. (2026). MSC Aging Agent Corpus (Version corpus-v2026.06.6) [Data set]. Syndicate Laboratories. https://huggingface.co/datasets/S4MPL3BI4S/msc-aging-agent-corpus
- 生物学主要来源应引用原始GEO或PubMed(通过
source_url或source_study_url),不要将Hugging Face作为生物学主要来源引用。
使用示例
python from huggingface_hub import hf_hub_download import json
REPO = "S4MPL3BI4S/msc-aging-agent-corpus" REV = "corpus-v2026.06.6"
provenance = hf_hub_download( repo_id=REPO, filename="datasets/CORPUS_PROVENANCE.json", repo_type="dataset", revision=REV, ) meta = json.load(open(provenance))
其他资源
- 构建源码与脚本: SampleBias/MSC_Age_Research
- 智能体入口文档: AGENT_CONNECT.md




