Gedik Huseyin, Peterson Roseann, Chatzinakos Christos, Dozmorov Mikhail G, Vladimirov Vladimir, Riley Brien P, Bacanu Silviu-Alin
medRxiv. 2024 Apr 15:2024.04.14.24305811. doi: 10.1101/2024.04.14.24305811.
Genome-wide association studies (GWAS) of psychiatric disorders (PD) yield numerous loci with significant signals, but often do not implicate specific genes. Because GWAS risk loci are enriched in expression/protein/methylation quantitative loci (e/p/mQTL, hereafter xQTL), transcriptome/proteome/methylome-wide association studies (T/P/MWAS, hereafter XWAS) that integrate xQTL and GWAS information, can link GWAS signals to effects on specific genes. To further increase detection power, gene signals are aggregated within relevant gene sets (GS) by performing gene set enrichment (GSE) analyses. Often GSE methods test for enrichment of "signal" genes in curated GS while overlooking their linkage disequilibrium (LD) structure, allowing for the possibility of increased false positive rates. Moreover, no GSE tool uses xQTL information to perform mendelian randomization (MR) analysis. To make causal inference on association between PD and GS, we develop a novel MR GSE (MR-GSE) procedure. First, we generate a "synthetic" GWAS for each MSigDB GS by aggregating summary statistics for x-level (mRNA, protein or DNA methylation (DNAm) levels) from the largest xQTL studies available) of genes in a GS. Second, we use synthetic GS GWAS as exposure in a generalized summary-data-based-MR analysis of complex trait outcomes. We applied MR-GSE to GWAS of nine important PD. When applied to the underpowered opioid use disorder GWAS, none of the four analyses yielded any signals, which suggests a good control of false positive rates. For other PD, MR-GSE greatly increased the detection of GO terms signals (2,594) when compared to the commonly used (non-MR) GSE method (286). Some of the findings might be easier to adapt for treatment, e.g., our analyses suggest modest positive effects for supplementation with certain vitamins and/or omega-3 for schizophrenia, bipolar and major depression disorder patients. Similar to other MR methods, when applying MR-GSE researchers should be mindful of the confounding effects of horizontal pleiotropy on statistical inference.
精神疾病（PD）的全基因组关联研究（GWAS）产生了许多具有显著信号的基因座，但通常无法确定具体的基因。由于GWAS风险基因座在表达/蛋白质/甲基化定量基因座（e/p/mQTL，以下简称xQTL）中富集，整合xQTL和GWAS信息的转录组/蛋白质组/甲基化组全基因组关联研究（T/P/MWAS，以下简称XWAS）可以将GWAS信号与对特定基因的影响联系起来。为了进一步提高检测能力，通过进行基因集富集（GSE）分析，将基因信号聚集在相关基因集（GS）内。通常，GSE方法会在经过整理的GS中测试“信号”基因的富集情况，而忽略它们的连锁不平衡（LD）结构，从而可能导致假阳性率增加。此外，没有GSE工具使用xQTL信息进行孟德尔随机化（MR）分析。为了对PD与GS之间的关联进行因果推断，我们开发了一种新颖的MR GSE（MR-GSE）程序。首先，我们通过汇总来自可用的最大xQTL研究的基因的x水平（mRNA、蛋白质或DNA甲基化（DNAm）水平）的汇总统计数据，为每个MSigDB GS生成一个“合成”GWAS。其次，我们在基于广义汇总数据的复杂性状结果的MR分析中，将合成GS GWAS用作暴露因素。我们将MR-GSE应用于九种重要PD的GWAS。当应用于功效不足的阿片类药物使用障碍GWAS时，四项分析均未产生任何信号，这表明对假阳性率有良好的控制。对于其他PD，与常用的（非MR）GSE方法（产生286个信号）相比，MR-GSE大大增加了GO术语信号的检测数量（2594个）。一些研究结果可能更容易应用于治疗，例如，我们的分析表明，补充某些维生素和/或ω-3对精神分裂症、双相情感障碍和重度抑郁症患者有适度的积极作用。与其他MR方法类似，应用MR-GSE时，研究人员应注意水平多效性对统计推断的混杂影响。
