Auvergne Antoine, Traut Nicolas, Henches Léo, Troubat Lucie, Frouin Arthur, Boetto Christophe, Kazem Sayeh, Julienne Hanna, Toro Roberto, Aschard Hugues
Department of Computational Biology, Institut Pasteur, Université Paris Cité, Paris, France.
Department of Computational Biology, Institut Pasteur, Université Paris Cité, Paris, France.
Biol Psychiatry Cogn Neurosci Neuroimaging. 2025 Jul;10(7):740-749. doi: 10.1016/j.bpsc.2024.08.018. Epub 2024 Sep 10.
There is increasing evidence of shared genetic factors between psychiatric disorders and brain magnetic resonance imaging (MRI) phenotypes. However, deciphering the joint genetic architecture of these outcomes has proven to be challenging, and new approaches are needed to infer the genetic structures that may underlie those phenotypes. Multivariate analyses are a meaningful approach to reveal links between MRI phenotypes and psychiatric disorders missed by univariate approaches.
First, we conducted univariate and multivariate genome-wide association studies for 9 MRI-derived brain volume phenotypes in 20,000 UK Biobank participants. Next, we performed various complementary enrichment analyses to assess whether and how univariate and multitrait approaches could distinguish disorder-associated and non-disorder-associated variants from 6 psychiatric disorders: bipolar disorder, attention-deficit/hyperactivity disorder, autism, schizophrenia, obsessive-compulsive disorder, and major depressive disorder. Finally, we conducted a clustering analysis of top associated variants based on their MRI multitrait association using an optimized k-medoids approach.
A univariate MRI genome-wide association study revealed only negligible genetic correlations with psychiatric disorders, while a multitrait genome-wide association study identified multiple new associations and showed significant enrichment for variants related to both attention-deficit/hyperactivity disorder and schizophrenia. Clustering analyses also detected 2 clusters that showed not only enrichment for association with attention-deficit/hyperactivity disorder and schizophrenia but also a consistent direction of effects. Functional annotation analyses of those clusters pointed to multiple potential mechanisms, suggesting in particular a role of neurotrophin pathways in both MRI phenotypes and schizophrenia.
Our results show that multitrait association signature can be used to infer genetically driven latent MRI variables associated with psychiatric disorders, thereby opening paths for future biomarker development.
越来越多的证据表明精神疾病与脑磁共振成像（MRI）表型之间存在共同的遗传因素。然而，解读这些结果的联合遗传结构已被证明具有挑战性，需要新的方法来推断可能构成这些表型基础的遗传结构。多变量分析是一种有意义的方法，可揭示单变量方法遗漏的MRI表型与精神疾病之间的联系。
首先，我们对20000名英国生物银行参与者的9种MRI衍生脑容量表型进行了单变量和多变量全基因组关联研究。接下来，我们进行了各种补充性富集分析，以评估单变量和多性状方法能否以及如何区分与6种精神疾病相关和不相关的变异：双相情感障碍、注意力缺陷多动障碍、自闭症、精神分裂症、强迫症和重度抑郁症。最后，我们使用优化的k-中心点方法，基于其MRI多性状关联对顶级相关变异进行聚类分析。
单变量MRI全基因组关联研究仅发现与精神疾病的遗传相关性可忽略不计，而多性状全基因组关联研究确定了多个新关联，并显示与注意力缺陷多动障碍和精神分裂症相关的变异有显著富集。聚类分析还检测到2个聚类，不仅显示与注意力缺陷多动障碍和精神分裂症的关联富集，而且效应方向一致。对这些聚类的功能注释分析指出了多种潜在机制，尤其表明神经营养因子通路在MRI表型和精神分裂症中均起作用。
我们的结果表明，多性状关联特征可用于推断与精神疾病相关的遗传驱动潜在MRI变量，从而为未来生物标志物的开发开辟道路。
