Endophenotype-based polygenic risk scores: Prediction of biomarker and clinical progression and dementia

Abstract BACKGROUND: Biomarkers provide a framework for a biological diagnosis of Alzheimer’s disease (AD) whereas polygenic risk scores (PRS) provide method to estimate genetic risk. We derive biomarker-based PRS by incorporating endophenotype genetic risk relevant to amyloid, tau, neurodegeneration and cerebrovascular (A/T/N/V) pathology. METHODS: Endophenotype-PRSs (PRSA, PRST, PRSN, PRSV) and combined-PRSs (PRSAT, PRSATNV) were generated using the Alzheimer’s Disease Neuroimaging Initiative (ADNI) data. Prediction performance of the PRSs was assessed in terms of dementia risk, age at onset (AAO) and longitudinal change of 14 important AD biomarkers. RESULTS: PRSA and PRST explained more amyloid and tau variability than combined PRSs (CSF-amyloid: R2PRSA = 9.22%; CSF-tau: R2PRST = 6.37%; CSF-ptau: R2PRST = 7.10%). Combined-PRSs explained more neurodegeneration-related variability (R2PRSATNV range: 1.22%-4.20%) and were strong predictors of dementia risk (HR and OR p-value<8.3e-03) and AAO (AAO(predicted_vs_observed): rAT=0.76). CONCLUSIONS: PRSA and PRST are AD-specific, while combined-PRSs are linked to neurodegeneration in general. Biomarker-derived PRSs provide mechanistic insights beyond aggregate disease susceptibility, supporting development of precision medicine for dementia.