Motor phenotypes: Multivariate associations with sleep, mental health, and grey matter volume

2057 

Abstract Submission 

Vincent Küppers^1,2, Vera Komeyer^2,3,4, Martin Gell^5,2, Federico Raimondo^2,4, Kaustubh Patil^2,4, Alexander Drzezga^1,6,2, Simon Eickhoff^2,4, Masoud Tahmasian^2,1,4

¹Department of Nuclear Medicine, University Hospital and Medical Faculty, University of Cologne, Cologne, Germany, ²Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich, Jülich, Germany, ³Department of Biology, Faculty of Mathematics and Natural Sciences, Heinrich Heine University Düsseldorf, Düsseldorf, Germany, ⁴Institute of Systems Neuroscience, Medical Faculty and University Hospital Düsseldorf, Heinrich Heine University Düsseldorf, Düsseldorf, Germany, ⁵Department of Psychiatry, Psychotherapy, and Psychosomatics, RWTH Aachen University Hospital, Aachen, Germany, ⁶German Center for Neurodegenerative Diseases (DZNE), Bonn-Cologne, Germany

Vincent Küppers  
Department of Nuclear Medicine, University Hospital and Medical Faculty, University of Cologne|Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich
Cologne, Germany|Jülich, Germany

Vera Komeyer  
Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich|Department of Biology, Faculty of Mathematics and Natural Sciences, Heinrich Heine University Düsseldorf|Institute of Systems Neuroscience, Medical Faculty and University Hospital Düsseldorf, Heinrich Heine University Düsseldorf
Jülich, Germany|Düsseldorf, Germany|Düsseldorf, Germany

Martin Gell  
Department of Psychiatry, Psychotherapy, and Psychosomatics, RWTH Aachen University Hospital|Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich
Aachen, Germany|Jülich, Germany

Federico Raimondo  
Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich|Institute of Systems Neuroscience, Medical Faculty and University Hospital Düsseldorf, Heinrich Heine University Düsseldorf
Jülich, Germany|Düsseldorf, Germany

Kaustubh Patil  
Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich|Institute of Systems Neuroscience, Medical Faculty and University Hospital Düsseldorf, Heinrich Heine University Düsseldorf
Jülich, Germany|Düsseldorf, Germany

Alexander Drzezga  
Department of Nuclear Medicine, University Hospital and Medical Faculty, University of Cologne|German Center for Neurodegenerative Diseases (DZNE)|Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich
Cologne, Germany|Bonn-Cologne, Germany|Jülich, Germany

Simon Eickhoff  
Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich|Institute of Systems Neuroscience, Medical Faculty and University Hospital Düsseldorf, Heinrich Heine University Düsseldorf
Jülich, Germany|Düsseldorf, Germany

Masoud Tahmasian  
Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich|Department of Nuclear Medicine, University Hospital and Medical Faculty, University of Cologne|Institute of Systems Neuroscience, Medical Faculty and University Hospital Düsseldorf, Heinrich Heine University Düsseldorf
Jülich, Germany|Cologne, Germany|Düsseldorf, Germany

Motor behaviour plays an important role in our ability to interact with the world. Higher levels of physical activity and motor performance are associated with better sleep quality and mental health [1], [2]. Previous studies mainly assessed univariate associations between individual measures rather than exploring the interplay between the latent dimensions of sleep, mental health, and motor behaviour. Furthermore, the neurobiology underlying their interplay remains unclear. In this study, we aimed to assess multivariate links between motor behaviour and the combined factors of sleep and mental health. Additionally, we discerned the interindividual neuroanatomical basis of their interaction through a predictive machine-learning approach in a large-scale sample.

We analysed data from 5853 participants (mean age=64, SD=7.7; 2907 female) without neurological diseases from the UK Biobank [3]. In total, 9 sleep health (e.g., insomnia symptoms) and 15 mental health (e.g., tiredness) variables were derived from touchscreen-based assessments [4], [5]. Motor behaviour was assessed using self-reported physical activity, accelerometer, grip strength, reaction time, and trail-making tasks. To prevent data leakage between the analyses, participants were split based on neuroimaging data availability. Regularized Canonical Correlation Analysis (rCCA) was applied to 1806 participants to identify multivariate associations between sleep/mental health (X) and motor behaviour (Y), validated through 5 repeated hold-out split and permutation tests [6]. The rCCA weights were then used to project the remaining 4047 participants, obtaining a motor latent variable for each participant and each rCCA mode. These variables were then used as targets for prediction with grey matter volume (GMV) data, parcellated using Schaefer 1000 ROIs, Melbourne Subcortical, and SUIT cerebellar atlases [7]-[9]. The 4047 participants were divided into 80/20 training and testing subsets. The training subset underwent nested 5-fold cross-validation using XGBoost. The final model was then retrained on the entire training set for out-of-sample predictions on the test subset.

Significant associations were identified between sleep/mental health and motor behaviour (p-omnibus: mode 1: 0.001, mode 2: 0.002), revealing two distinct association modes (Fig.1). The first mode showed higher grip strength and lower physical activity, which was positively associated with snoring, napping, and intermediate chronotype, and negatively associated with neuroticism, morning chronotype, and insomnia symptoms. The second mode highlighted higher self-reported physical activity and grip strength, associated with less depressive symptoms, less difficulties in awakening, and a tendency toward morning chronotype. The same modes were not observed after removing age and sex effects. Initially, predictive analyses of the motor latent variable of both modes showed moderate links to GMV (mode 1: R2=0.17, mode 2: R2=0.18; Fig.2). However, the predictability of mode 1 substantially dropped after linearly adjusting features for age and sex in the predictive analysis (R2=0.04, r=0.19) or when adjusting the rCCA for age and sex (R2=0.02, r=0.15). A similar pattern was observed for mode 2.

We found two distinct motor latent phenotypes intricately associated with sleep/mental health and modestly linked to GMV variability. The first phenotype, characterised by increased strength but low physical activity, correlated with lower neuroticism and increased snoring/napping. The second, indicating higher self-reported physical activity was linked to better sleep health and reduced depressive symptoms. These findings highlight potential links between sleep/mental health with motor behaviour on multiple levels. They further underscore the importance of a thorough evaluation of age and sex-related effects and show potential of exploring psychologically and biologically informed phenotypes for brain-based predictive models.

Classification and Predictive Modeling

Multivariate Approaches ²

Motor Behavior Other ¹

Cortical Anatomy and Brain Mapping

Sleep and Wakefulness

Machine Learning

Motor

Multivariate

Sleep

STRUCTURAL MRI

Other - Mental Health