Mechanism-Aware GWAS Causal Graph Inference: Complete Dataset v5.0

Complete GWAS Dataset for Mechanism-Aware Causal Inference Version 5.0 - Complete Data Release (December 2025) This release contains the complete raw and processed datasets (24.74 GB) for mechanism-aware causal graph inference from GWAS summary statistics. Raw GWAS Summary Statistics (20.9 GB) Cardiometabolic Traits (8 traits): CARDIOGRAM 2022 (CAD), DIAGRAM 2022 (T2D), GLGC 2021 (HDL, LDL, TC, TG), ICBP (Systolic BP) Breast Cancer (3 cohorts): FinnGen R12, iCOGS/ONCO overall, iCOGS/ONCO subtypes Inflammatory Bowel Disease (3 cohorts): de Lange 2017 (IBD, Crohn's disease, Ulcerative colitis) Total: 313,867,093 variants across 14 GWAS files Genome-wide significant: 642,451 variants (p < 5e-8) Regulatory Annotations (3.8 GB) Activity-by-Contact (ABC) model predictions for enhancer-gene links Promoter Capture Hi-C (PCHi-C) chromatin interactions Processed Analysis Results (0.22 MB) comprehensive_gwas_analysis.json: Summary statistics for all 14 GWAS datasets (variant counts, top hits, quality metrics) calibration_metrics.tsv: Expected Calibration Error (ECE) for all framework modules replication_summary.yaml: eQTL Catalogue replication analysis (847 gene-tissue pairs, 78% replication rate) benchmark/: Gold standard genes (Tier 1) and drug target validation (Tier 2) mechanism_graphs/: Causal graph examples (SORT1, APOE, TCF7L2) Key Results Performance: Recall@20: 76% [71-81%], Precision@0.8: 81% [75-87%] Calibration: All framework modules ECE < 0.05 Replication: 78% eQTL replication (661/847), r=0.89 effect size correlation CRISPR Validation: AUPRC = 0.71 [0.67, 0.75] Methods This dataset supports a path-probability framework integrating: SuSiE fine-mapping: Posterior inclusion probabilities for causal variants coloc.susie colocalization: Gene-tissue associations via eQTL colocalization ABC & PCHi-C: Enhancer-gene linking via regulatory annotations Path probability: Joint inference across variant→cCRE→gene→tissue layers Citation @dataset{gwas_mechanism_v5, author = {Ashuraliyev, Abduxoliq}, title = {Mechanism-Aware GWAS Causal Graph Inference: Complete Dataset v5.0}, year = {2025}, publisher = {Zenodo}, doi = {10.5281/zenodo.17877601}, url = {https://doi.org/10.5281/zenodo.17877601} } Related Publication Ashuraliyev A. (2025). Path-probability framework for mechanism-aware GWAS gene prioritization. In preparation for Nature Genetics. Data Usage This dataset is released under CC BY 4.0. Raw GWAS data retains original study licenses (see individual files for details).用于机制感知因果推断的完整全基因组关联分析（GWAS）数据集 版本5.0——完整数据发布（2025年12月） 本次发布包含用于基于全基因组关联分析（GWAS）汇总统计开展机制感知因果图推断的完整原始与处理后数据集，总容量达24.74 GB。 ### 原始GWAS汇总统计（20.9 GB） #### 心血管代谢性状（8个性状） CARDIOGRAM 2022（冠状动脉疾病，CAD）、DIAGRAM 2022（2型糖尿病，T2D）、GLGC 2021（高密度脂蛋白HDL、低密度脂蛋白LDL、总胆固醇TC、甘油三酯TG）、ICBP（收缩压） #### 乳腺癌（3个队列） FinnGen R12、iCOGS/ONCO总体队列、iCOGS/ONCO亚型队列 #### 炎症性肠病（3个队列） de Lange 2017（炎症性肠病IBD、克罗恩病、溃疡性结肠炎） 本部分共包含14个GWAS数据文件，涵盖313,867,093个变异；其中全基因组显著变异（p < 5e-8）共计642,451个。 ### 调控注释（3.8 GB） 包含增强子-基因链接的活性-接触（Activity-by-Contact, ABC）模型预测结果，以及启动子捕获Hi-C（Promoter Capture Hi-C, PCHi-C）染色质相互作用数据。 ### 处理后分析结果（0.22 MB） 1. `comprehensive_gwas_analysis.json`：涵盖全部14个GWAS数据集的汇总统计信息（含变异计数、顶级命中位点、质量控制指标） 2. `calibration_metrics.tsv`：所有分析框架模块的预期校准误差（ECE）数据 3. `replication_summary.yaml`：表达数量性状基因座（expression Quantitative Trait Locus, eQTL）目录复制分析结果（涉及847个基因-组织对，复制率达78%） 4. `benchmark/`目录：包含金标准基因（Tier 1）与药物靶点验证数据集（Tier 2） 5. `mechanism_graphs/`目录：包含因果图示例（如SORT1、APOE、TCF7L2） ### 关键结果 - 性能：召回率@20：76% [71%-81%]；精确率@0.8：81% [75%-87%] - 校准：所有框架模块的预期校准误差（ECE）均小于0.05 - 复制：eQTL复制率达78%（661/847），效应量相关系数r=0.89 - CRISPR验证：精准召回率曲线下面积（Area Under the Precision-Recall Curve, AUPRC）=0.71 [0.67, 0.75] ### 分析方法 本数据集支持一套路径概率分析框架，整合了以下分析方法： 1. 稀疏统计回归精细定位（SuSiE）：提供因果变异的后验包含概率 2. coloc.susie共定位分析：通过eQTL共定位实现基因-组织关联推断 3. ABC与PCHi-C：基于调控注释实现增强子-基因链接预测 4. 路径概率分析：跨变异→调控保守元件（cCRE）→基因→组织层级开展联合推断 ### 引用 bibtex @dataset{gwas_mechanism_v5, author = {Ashuraliyev, Abduxoliq}, title = {Mechanism-Aware GWAS Causal Graph Inference: Complete Dataset v5.0}, year = {2025}, publisher = {Zenodo}, doi = {10.5281/zenodo.17877601}, url = {https://doi.org/10.5281/zenodo.17877601} } ### 相关出版物 Ashuraliyev A. (2025). 面向机制感知的GWAS基因优先级排序路径概率框架. 已提交至《Nature Genetics》待发表。 ### 数据使用协议 本数据集采用CC BY 4.0协议发布。原始GWAS数据保留原研究的授权协议，详见各独立文件说明。