Machine learning approach to DNA methylation and neuroimaging signatures as biomarkers for psychological resilience in young adults.
Authors
Affiliations (8)
Affiliations (8)
- Department of Psychiatry, National Cheng Kung University Hospital, College of Medicine, National Cheng Kung University, Tainan, Taiwan; Institute of Behavioral Medicine, College of Medicine, National Cheng Kung University, Tainan, Taiwan; Clinical Medicine Research Center, National Cheng Kung University Hospital, College of Medicine, National Cheng Kung University, Tainan, Taiwan.
- Department of Public Health, College of Medicine, National Cheng Kung University, Tainan, Taiwan.
- Department of Psychology, College of Social Sciences, National Cheng Kung University, Tainan, Taiwan.
- Institute of Clinical Medicine, College of Medicine, National Cheng Kung University, Tainan, Taiwan.
- Department of Psychology, College of Social Sciences, National Cheng Kung University, Tainan, Taiwan; Institute of Allied Health Sciences, College of Medicine, National Cheng Kung University, Tainan, Taiwan; Graduate Institute of Health and Biotechnology Law, Taipei Medical University, Taipei, Taiwan.
- Department of Psychology, College of Social Sciences, National Cheng Kung University, Tainan, Taiwan; Institute of Gerontology, College of Medicine, National Cheng Kung University, Tainan, Taiwan; Institute of Behavioral Medicine, College of Medicine, National Cheng Kung University, Tainan, Taiwan; Institute of Genomics and Bioinformatics, College of Life Sciences, National Chung Hsing University, Taichung, Taiwan.
- Department of Public Health, College of Medicine, National Cheng Kung University, Tainan, Taiwan; Department of Psychology, College of Social Sciences, National Cheng Kung University, Tainan, Taiwan; Institute of Allied Health Sciences, College of Medicine, National Cheng Kung University, Tainan, Taiwan. Electronic address: [email protected].
- Department of Public Health, College of Medicine, National Cheng Kung University, Tainan, Taiwan; Institute of Clinical Medicine, College of Medicine, National Cheng Kung University, Tainan, Taiwan; Biostatistics Consulting Center, National Cheng Kung University Hospital, College of Medicine, National Cheng Kung University, Tainan, Taiwan. Electronic address: [email protected].
Abstract
Psychological resilience is influenced by both psychological and biological factors. However, the potential of using DNA methylation (DNAm) probes and brain imaging variables to predict psychological resilience remains unclear. This study aimed to investigate DNAm, structural magnetic resonance imaging (sMRI), and diffusion tensor imaging (DTI) as biomarkers for psychological resilience. Additionally, we evaluated the ability of epigenetic and imaging markers to distinguish between individuals with low and high resilience using machine learning algorithms. A total of 130 young adults assessed with the Connor-Davidson Resilience Scale (CD-RISC) were divided into high and low psychological resilience groups. We utilized two feature selection algorithms, the Boruta and variable selection using random forest (varSelRF), to identify important variables based on nine for DNAm, sixty-eight for gray matter volume (GMV) measured with sMRI, and fifty-four diffusion indices of DTI. We constructed machine learning models to identify low resilience individuals using the selected variables. The study identified thirteen variables (five DNAm, five GMV, and three DTI diffusion indices) from feature selection methods. We utilized the selected variables based on 10-fold cross validation using four machine learning models for low resilience (AUC = 0.77-0.82). In interaction analysis, we identified cg03013609 had a stronger interaction with cg17682313 and the rostral middle frontal gyrus in the right hemisphere for psychological resilience. Our findings supported the concept that DNAm, sMRI, and DTI signatures can identify individuals with low psychological resilience. These combined epigenetic imaging markers demonstrated high discriminative abilities for low psychological resilience using machine learning models.