Genetic architecture of MRI-derived intervertebral DISC degeneration revealed by automated phenotyping and genomic structural equation modelling.
Authors
Affiliations (7)
Affiliations (7)
- Department of Twin Research and Genetic Epidemiology, School of Life Course and Population Sciences, King's College London, London, UK. [email protected].
- Department of Twin Research and Genetic Epidemiology, School of Life Course and Population Sciences, King's College London, London, UK.
- School of Medicine, University of Limerick, Limerick, Ireland.
- Research Unit of Health Sciences and Technology, University of Oulu, Oulu, Finland.
- Wellbeing Services County of South Karelia, Lappeenranta, Finland.
- Department of Anaesthesiology, Michigan Medicine, Ann Arbor, USA.
- Michigan Genomics Initiative, AI & Digital Health Innovation, University of Michigan-Ann Arbor, Ann Arbor, USA.
Abstract
Intervertebral disc degeneration (IVDD) encompasses structural changes of the spine that are commonly studied using clinically defined diagnoses or symptom-based phenotypes, which reflect a mixture of anatomical degeneration and pain-related processes. Magnetic resonance imaging (MRI) provides direct assessment of spinal structure, but genetic studies using imaging-derived traits have been limited by challenges in large-scale phenotyping. We applied SpineNet, a deep learning framework for automated MRI grading, to derive lumbar spine degeneration phenotypes in 7,414 individuals from 3 cohorts. GWAS was conducted within each cohort and results meta-analysed. Phenotypic and genetic relationships were assessed using exploratory factor analysis and LD score regression. genomicSEM modelled shared genetic liability across MRI traits and with GWAS of chronic pain and spinal disorders, followed by a common factor GWAS. GenomicSEM identified a latent genetic dimension representing structural spinal degeneration, with strong contributions from disc narrowing, endplate defect, herniation, foraminal stenosis, and marrow signal change. When integrated with pain and clinical phenotypes, structural MRI traits and chronic pain had distinct genetic dimensions, while clinically defined spinal diagnoses showed contributions from both. The common factor GWAS identified several suggestive loci enriched in regulatory genomic regions, mapping to genes previously implicated in musculoskeletal traits. Phenotypic and genetic analyses revealed a consistent factor structure in which MRI-derived structural traits clustered together and were distinct from alignment-related abnormalities. Automated MRI phenotyping combined with multivariate genetic modelling identified a genetic dimension of structural spinal degeneration that is partially distinct from pain-related liability, providing a scalable framework for future spine studies.