基于孟德尔随机化的基因-代谢物-胰腺癌风险因果链解析

Causal chain linking genes, metabolites, and pancreatic cancer risk based on Mendelian randomization

摘要:

背景与目的 胰腺癌（PC）是预后极差的恶性肿瘤，早期诊断与风险预测仍面临巨大挑战。既往研究提示多种血清代谢物与PC风险相关，但因观察性研究受混杂因素影响，其因果性尚未明确。本研究基于大规模基因组数据，系统评估血清代谢物与PC风险的遗传因果关系，筛选PC风险相关基因，并构建基因-代谢物-PC的因果关联网络。方法 利用双样本孟德尔随机化（TSMR）和基于汇总数据的孟德尔随机化（SMR），分析方法，整合325种血清代谢物的GWAS数据、胰腺组织eQTL数据及FinnGen R12 PC GWAS数据，分析代谢物与PC风险的遗传因果关联，筛选PC风险基因，并进一步探讨代谢物在基因与PC风险间的潜在中介作用。结果 TSMR分析共鉴定出4种血清代谢物与PC风险存在显著因果关系。血清白蛋白水平（OR=1.456，P=0.003）及小高密度脂蛋白（HDL）中游离胆固醇比例（OR=1.189，P=0.005）升高显著增加PC风险，而中间密度脂蛋白（IDL）与小HDL中磷脂比例升高则具有保护作用（OR=0.792和0.836，均P<0.01）。SMR分析进一步识别出196个与PC风险显著相关的基因，其中包括DGKQ、CDC37P1、SULT1A2等风险基因及PALMD、HEG1等保护性基因。最终共确立32对显著的基因-代谢物因果配对，揭示特定代谢物在基因与PC风险之间可能发挥中介作用。结论 本研究系统描绘了血清代谢物与胰腺癌风险之间的因果关联及其基因调控网络，提示脂质代谢相关通路在PC发生中的核心作用。所得代谢物及基因线索为胰腺癌的早期诊断、分层筛查及个体化干预策略提供了科学依据。

Abstract:

Background and Aims Pancreatic cancer (PC) is a highly lethal malignancy with poor prognosis and limited early diagnostic tools. Although numerous serum metabolites have been associated with PC risk in observational studies, the causal nature of these associations remains uncertain. This study aimed to evaluate the genetic causal relationships between serum metabolites and PC risk, identify PC-related risk genes, and elucidate the gene-metabolite-PC causal network.Methods Two-sample Mendelian randomization (TSMR) and summary-data-based Mendelian randomization (SMR) analyses were performed by integrating GWAS data of 325 serum metabolites, GTEx v8 pancreatic tissue eQTL data, and FinnGen R12 PC GWAS data. The study assessed causal effects of metabolites on PC risk, identified risk-associated genes, and explored the potential mediating role of metabolites between genes and PC.Results Four serum metabolites showed significant causal relationships with PC risk. Elevated serum albumin (OR=1.456, P=0.003) and free cholesterol percentage in small high-density lipoprotein (HDL) (OR=1.189, P=0.005) were associated with increased PC risk, whereas higher phospholipid percentages in intermediate-density lipoprotein (IDL) and small HDL were protective (OR=0.792 and 0.836, respectively; both P<0.01). SMR analysis identified 196 PC-related genes, including risk genes such as DGKQ, CDC37P1, and SULT1A2, and protective genes such as PALMD and HEG1. Thirty-two significant gene-metabolite causal pairs were further confirmed, indicating potential mediation of PC genetic risk through specific metabolic pathways.Conclusion This study systematically clarified the causal relationships between serum metabolites and pancreatic cancer risk and established a gene-metabolite regulatory network. The findings highlight the central role of lipid metabolism in PC development and provide molecular evidence for early detection and personalized prevention strategies.