Download This Paper
Preprints with The Lancet is part of SSRN´s First Look, a place where journals identify content of interest prior to publication. Authors have opted in at submission to The Lancet family of journals to post their preprints on Preprints with The Lancet. The usual SSRN checks and a Lancet-specific check for appropriateness and transparency have been applied. Preprints available here are not Lancet publications or necessarily under review with a Lancet journal. These preprints are early stage research papers that have not been peer-reviewed. The findings should not be used for clinical or public health decision making and should not be presented to a lay audience without highlighting that they are preliminary and have not been peer-reviewed. For more information on this collaboration, see the comments published in The Lancet about the trial period, and our decision to make this a permanent offering, or visit The Lancet´s FAQ page, and for any feedback please contact preprints@lancet.com.
Multimodality-Based Graph Convolutional Networks for Classification of Adults with Attention-Deficit/Hyperactivity Disorder and Detection of Potential Neuroimaging Genetics Biomarkers
42 Pages Posted: 28 Aug 2024
More...Abstract
Background: Intergrating multimodal data across neuroimaging genetics can enhance the diagnosis and understanding of adults with Attention-deficit/hyperactivity disorder (aADHD). However, most studies only focused on classification via machine learning (ML), lacking detailed explanations of unimodal biomarkers or relationships between the multimodal data. Here, we employed a population-based Graph Convolutional Network (GCN) model and conducted a comprehensive downstream analysis to uncover the pathophysiology of aADHD.
Methods: Resting-state functional MRI and genomics data were collected from 247 aADHD patients and 237 healthy controls. The multimodal data were combined to construct a population graph, which was trained using a semi-supervised Edge-Variational GCN (EV-GCN) model. Through sensitivity analysis, salient functional connections (FCs) and single nucleotide polymorphisms (SNPs) contributing to aADHD were identified and used to calculate correlations with clinical traits. Moreover, downstream association analyses were conducted to explore the genetic components linked to salient imaging alterations.
Findings: Our model achieved an accuracy of 73.55% with standard deviation of 6.79% (AUC: 0.791), which exceeded existing classifiers. The most salient FCs were mainly located in the dorsal attention, limbic, and default mode networks. Among these, the FC between the right precuneus (PCun) and left parahippocampal gyrus was significantly correlated with the ADHD core symptoms. Key hubs, particularly the right PCun, were identified from the top salient FCs, and their topological properties were associated with hyperactivity/impulsivity and executive function performance. Notably, the deficient FC of the right PCun exhibited associations with the genetic variant of HDAC9 in genome-wide association studies, as well as with the expression profiles of COL19A1 and CSMD2. These latter two genes were coded by the salient SNPs identified by our classifier.
Interpretation: These findings highlight the effectiveness and interpretability of the multimodality-based EV-GCN in discriminating aADHD individuals from healthy controls. The functional alterations in right PCun and its related genetic profiles were suggested to play an important role in the neurological underpinnings of aADHD.
Funding: The study was supported by the National Natural Science Foundation of China (82271575, 62273010), Beijing Nova Program (20220484061; 20230484444), Capital’s Funds for Health Improvement and Research (CFH: 2024-2-4114),Clinical Medicine Plus X—Young Scholars Project, Peking University, the Fundamental Research Funds for the Central Universities (PKU2023LCXQ043), and Beijing Municipal Health Commission Research Ward Programme (3rd batch). This research is supported by high-performance computing platform of Peking University.
Declaration of Interest: No potential conflict of interest was reported by the author(s).
Ethical Approval: Informed consent was obtained from all participants, and experimental procedure was approved by the Research Ethics Review Board of Peking University Sixth Hospital.
Keywords: Adult With Attention-Deficit/Hyperactivity Disorder, Neuroimaging Genetics, Deep Learning, Population-based Graph Convolution Network
Suggested Citation: Suggested Citation