Development of a Multi-Trait Polygenic Score for Intrinsic Capacity

Background: Intrinsic capacity (IC) is a key marker of healthy ageing, which captures an individuals physical and mental capacities, measured across five domains: cognitive, locomotor, psychological, vitality, and sensory. Although genetic factors are known to influence both general IC and its individual domains, existing IC indices have been developed primarily using phenotypic data, without accounting for the underlying biological architecture across domains. In this study, we developed a multi-trait polygenic score (Mt-PGS) model for IC by integrating polygenic scores derived from a broad set of phenotypes spanning the five IC domains and examined its validity. Methods: Using data from 13,085 participants of the Canadian Longitudinal Study on Aging (CLSA), we computed PGSs for 63 phenotypes related to IC domains. A supervised machine-learning model was applied to develop a mt-PGS model for IC and identify the optimal set of polygenic predictors. The validity of the mt-PGS IC score was evaluated by comparing it with a phenotype-based IC score and by examining its association with mortality. Results: Our analysis identified PGSs for 33 phenotypes with non-zero coefficients, jointly explaining 2.23% of the variance in IC. Several of the strongest contributors were most closely aligned with vitality-related phenotypes in the literature (including body mass index, grip strength, fat-free mass, diastolic blood pressure, and chronic obstructive pulmonary disease), acknowledging cross-domain relevance, and that predictors from all five IC domains were represented. The mt-PGS IC score was consistent with the phenotype-based IC score, positively correlated with the phenotype-based IC score and was inversely associated with mortality (OR = 0.04; 95% CI: 0.005 - 0.379). Conclusion: Our findings support the multisystem biological basis of IC, demonstrating that an mt-PGS model integrating diverse phenotypes is associated with the phenotype-based IC score. PGSs for the phenotypes frequently related to vitality in the literature were the strongest predictors, recognizing that several of these phenotypes may span multiple domains, and that all domains contributed to the model. If replicated across different ancestries and settings, these findings may serve as a foundation for future research for the potential integration of genetic information into IC frameworks. Keywords: Intrinsic capacity, Polygenic scores, CLSA, Machine learning, Healthy ageing