Machine learning suggests polygenic risk for cognitive dysfunction in amyotrophic lateral sclerosis

Amyotrophic lateral sclerosis (ALS) is a multi‐system disease characterized primarily by progressive muscle weakness. Cognitive dysfunction is commonly observed in patients; however, factors influencing risk for cognitive dysfunction remain elusive. Using sparse canonical correlation analysis (sCCA), an unsupervised machine‐learning technique, we observed that single nucleotide polymorphisms collectively associate with baseline cognitive performance in a large ALS patient cohort ( N = 327) from the multicenter Clinical Research in ALS and Related Disorders for Therapeutic Development (CReATe) Consortium. We demonstrate that a polygenic risk score derived using sCCA relates to longitudinal cognitive decline in the same cohort and also to in vivo cortical thinning in the orbital frontal cortex, anterior cingulate cortex, lateral temporal cortex, premotor cortex, and hippocampus ( N = 90) as well as post‐mortem motor cortical neuronal loss ( N = 87) in independent ALS cohorts from the University of Pennsylvania Integrated Neurodegenerative Disease Biobank. Our findings suggest that common genetic polymorphisms may exert a polygenic contribution to the risk of cortical disease vulnerability and cognitive dysfunction in ALS. Synopsis Single nucleotide polymorphisms (SNPs) previously identified through genome‐wide association studies as risk factors for amyotrophic lateral sclerosis (ALS) and/or frontotemporal dementia (FTD) further associate in a polygenic manner with risk for cognitive dysfunction in ALS and related disorders. A machine learning approach identified a subset of SNPs that maximally contributed to variance in cognitive performance in a large patient cohort. A polygenic risk score using the subset of SNPs and weights from machine learning analyses related to 1) in vivo cortical thinning in the frontal and temporal lobes, and 2) post mortem neuronal loss in the motor cortex in independent patient cohorts. Sparse Canonical Correlation Analysis (sCCA) provides an alternative approach to polygenic risk score generation. Graphical Abstract Single nucleotide polymorphisms (SNPs) previously identified through genome‐wide association studies as risk factors for amyotrophic lateral sclerosis (ALS) and/or frontotemporal dementia (FTD) further associate in a polygenic manner with risk for cognitive dysfunction in ALS and related disorders.