Using a polygenic score in a family design to understand genetic influences on musicality

Laura W. Wesseldijk, Abdel Abdellaoui, Reyna L. Gordon, Stella Aslibekyan, Adam Auton, Elizabeth Babalola, Robert K. Bell, Jessica Bielenberg, Katarzyna Bryc, Emily Bullis, Daniella Coker, Gabriel Cuéllar-Partida, Devika Dhamija, Sayantan Das, Sarah L. Elson, Teresa Filshtein, Kipper Fletez‐Brant, Pierre Fontanillas, Will Freyman, Anna Faaborg, Shirin T. Fuller, Pooja Gandhi, Karl Heilbron, Barry Hicks, Ethan M. Jewett, Katelyn Kukar, Keng‐Han Lin, Maya Lowe, Jey C. McCreight, Matthew H. McIntyre, Steven J. Micheletti, Meghan E. Moreno, Joanna L. Mountain, Priyanka Nandakumar, Elizabeth S. Noblin, Jared O’Connell, Yunru Huang, Aaron A. Petrakovitz, Vanessa Lane, Joanne S. Kim, G. David Poznik, Morgan Schumacher, Anjali J. Shastri, Janie F. Shelton, Jingchunzi Shi, Suyash Shringarpure, Vinh Tran, Joyce Y. Tung, Xin Wang, Wei Wang, Catherine H. Weldon, Peter Wilton, Alejandro Hernandez, Corinna Wong, Christophe Toukam Tchakouté, Fredrik Ullén, Miriam A. MosingScientific 2022 · DOI

TL;DR

This study uses a polygenic score (PGS) derived from a large-scale genetic study of beat synchronization to investigate genetic influences on musicality in Swedish twins. The findings suggest that this PGS captures a general genetic propensity for musicality, influencing various music-related skills and behaviors more directly than through indirect genetic or confounding pathways.

What Problem Does This Paper Address?

Understanding the genetic underpinnings of musicality is complex. While twin studies have established a significant genetic component to musical skills and engagement, identifying the specific genes involved has been challenging. Previous research has relied on candidate gene studies, which are prone to replication issues and small sample sizes. More recently, genome-wide association studies (GWAS) have identified genetic variants associated with specific musical traits, such as self-reported beat synchronization ability. However, these studies often capture only a fraction of the total genetic variation ("missing heritability"). Furthermore, it's crucial to disentangle direct genetic effects from indirect genetic influences (e.g., gene-environment correlations) and confounding factors (e.g., population stratification, assortative mating) that can inflate observed associations. This paper aims to address these challenges by using a polygenic score (PGS) approach in a family design to explore the genetic architecture of musicality and its potential pathways.

Methods

The core methodological innovation here is the application of a polygenic score (PGS) within a family design, specifically utilizing dizygotic (DZ) twins.

  • Polygenic Score (PGS) Construction: A PGS for rhythm ability (PGS<sub>rhythm</sub>) was created by aggregating the effect sizes of genetic variants identified in a large GWAS of self-reported beat synchronization ability. To improve accuracy, the effect sizes were re-estimated using a summary-data based best linear unbiased prediction (SBLUP) approach, accounting for linkage disequilibrium between SNPs.
  • Within-Family Design: The study leverages the unique structure of twin data. By comparing effects between unrelated individuals (or across families) and within DZ twin pairs (who share genes and environment), researchers can disentangle different pathways.
  • Between-Family (β<sub>B</sub>) effects: Reflect genetic and environmental influences shared by family members, as well as population-level effects.
  • Within-Family (β<sub>W</sub>) effects: Primarily capture direct genetic effects, as DZ twins share the same parents and environment but differ in their genetic makeup. If β<sub>W</sub> is significantly smaller than β<sub>B</sub>, it suggests indirect genetic or confounding effects are inflating the overall association.
  • Gene-Environment Correlation (rGE) Testing: The study also investigated potential gene-environment correlations by examining the relationship between an individual's PGS<sub>rhythm</sub> and their reported family musical enrichment environment.
  • Confounding Factor Assessment: Potential confounding from assortative mating and population structure was assessed by examining the correlation of PGS<sub>rhythm</sub> between DZ twins. A correlation significantly different from the expected 0.50 would indicate confounding.

Datasets

  • Swedish Twin Registry (STR): The primary dataset comprises 5648 Swedish twins from the Study of Twin Adults: Genes and Environment (STAGE) cohort. These individuals completed web surveys on musical engagement, aptitude, and other traits.
  • Genomic Data: Genotype data was available for 8343 individuals from the STAGE cohort, which was then imputed.
  • 23andMe Research Team Data: The GWAS summary statistics for self-reported beat synchronization ability were sourced from a large study involving participants from 23andMe, Inc.

Key Insights and Ideas

  • PGS<sub>rhythm</sub> as a General Musical Propensity Score: The study strongly suggests that the PGS derived from a simple beat synchronization task doesn't just predict rhythm ability but reflects a broader genetic predisposition for musicality. This is a key insight, moving beyond narrow trait associations.
  • Direct Genetic Effects Dominate: The within-family analyses provided compelling evidence that the PGS<sub>rhythm</sub> influences musical skills more directly, rather than being significantly inflated by indirect genetic effects (like passive gene-environment correlation) or confounding factors. This is a crucial finding for understanding how genes translate into behavior.
  • Gene-Environment Correlation is Present, but Not the Primary Driver: While a correlation between an individual's genetic propensity for rhythm and their family's musical environment was found, this correlation did not appear to inflate the direct genetic effects on musical skills within families. This suggests that while the environment might play a role, the genetic predisposition itself is a strong direct predictor.
  • Specificity of Genetic Influence: The PGS<sub>rhythm</sub> showed significant associations with various music-related traits (rhythm, melody, pitch discrimination, motor timing, practice, achievement) but not with control phenotypes like general intelligence or sport practice. This highlights the specificity of the genetic variants captured by the PGS.

Key Results

  • Predictive Power: The PGS<sub>rhythm</sub> significantly predicted rhythm discrimination (β=0.11, R<sup>2</sup>=0.01), melody discrimination (β=0.13, R<sup>2</sup>=0.02), pitch discrimination (β=0.13, R<sup>2</sup>=0.02), and overall musical aptitude (β=0.16, R<sup>2</sup>=0.03). It also predicted motor timing (β=0.17, R<sup>2</sup>=0.03) and other music-related outcomes.
  • Lack of Association with Control Phenotypes: The PGS<sub>rhythm</sub> did not predict general intelligence, total sport practice, or flow proneness/achievements in non-musical domains. A small, significant effect was observed for achievement in dance (β=0.04, R<sup>2</sup>=0.001).
  • Within-Family vs. Between-Family Effects: Within-family effects for rhythm discrimination and musical aptitude were not lower than between-family effects, and their confidence intervals overlapped. This indicates that indirect genetic or confounding effects did not significantly inflate the PGS<sub>rhythm</sub>'s influence.
  • Gene-Environment Correlation: A small but significant positive correlation (r=0.05, p<0.001) was found between musical enrichment of the family childhood environment and individuals' PGS<sub>rhythm</sub>.
  • Potential Confounding: The correlation between DZ twins for PGS<sub>rhythm</sub> was 0.54 (CI 0.50–0.58), which was borderline significant and included 0.50, suggesting that confounding effects like assortative mating or population structure were not strongly inflating the results.

Limitations and Future Work

  • Small Variance Explained: The authors acknowledge that the variance explained by the PGS<sub>rhythm</sub> for music outcomes is small, which is typical for PGS studies. They attribute this to the limitations of current GWAS sample sizes and the complexity of musical traits. Larger sample sizes and more objective, harmonized musical phenotyping across studies are needed to increase predictive accuracy.
  • Common Variants Only: The current PGS is based on common genetic variants. It's unknown whether rare genetic variants play a significant role.
  • Limited Power for Within-Family Analyses: The within-family analyses, particularly for disentangling indirect/confounding effects, were conducted on a smaller sample of 243 complete DZ twin pairs, which may have limited statistical power to detect subtle differences.
  • Future Directions: The authors suggest that this PGS<sub>rhythm</sub> can serve as a valuable tool for future research, particularly for meta-analyses and for disentangling gene-environment interplay in musicality. They emphasize the potential for future GWAS to identify more generalist music genes by increasing sample sizes and improving phenotyping.

Summary updated

Abstract

To further our understanding of the genetics of musicality, we explored associations between a polygenic score for self-reported beat synchronization ability (PGS<sub>rhythm</sub>) and objectively measured rhythm discrimination, as well as other validated music skills and music-related traits. Using family data, we were able to further explore potential pathways of direct genetic, indirect genetic (through passive gene-environment correlation) and confounding effects (such as population structure and assortative mating). In 5648 Swedish twins, we found PGS<sub>rhythm</sub> to predict not only rhythm discrimination, but also melody and pitch discrimination (betas between 0.11 and 0.16, p < 0.001), as well as other music-related outcomes (p < 0.05). In contrast, PGS<sub>rhythm</sub> was not associated with control phenotypes not directly related to music. Associations did not deteriorate within families (N = 243), implying that indirect genetic or confounding effects did not inflate PGS<sub>rhythm</sub> effects. A correlation (r = 0.05, p < 0.001) between musical enrichment of the family childhood environment and individuals' PGS<sub>rhythm</sub>, suggests gene-environment correlation. We conclude that the PGS<sub>rhythm</sub> captures individuals' general genetic musical propensity, affecting musical behavior more likely direct than through indirect or confounding effects.
BibTeX
@inproceedings{wesseldijk2022using,
  author = {Laura W. Wesseldijk and Abdel Abdellaoui and Reyna L. Gordon and Stella Aslibekyan and Adam Auton and Elizabeth Babalola and Robert K. Bell and Jessica Bielenberg and Katarzyna Bryc and Emily Bullis and Daniella Coker and Gabriel Cuéllar-Partida and Devika Dhamija and Sayantan Das and Sarah L. Elson and Teresa Filshtein and Kipper Fletez‐Brant and Pierre Fontanillas and Will Freyman and Anna Faaborg and Shirin T. Fuller and Pooja Gandhi and Karl Heilbron and Barry Hicks and Ethan M. Jewett and Katelyn Kukar and Keng‐Han Lin and Maya Lowe and Jey C. McCreight and Matthew H. McIntyre and Steven J. Micheletti and Meghan E. Moreno and Joanna L. Mountain and Priyanka Nandakumar and Elizabeth S. Noblin and Jared O’Connell and Yunru Huang and Aaron A. Petrakovitz and Vanessa Lane and Aaron A. Petrakovitz and Joanne S. Kim and G. David Poznik and Morgan Schumacher and Anjali J. Shastri and Janie F. Shelton and Jingchunzi Shi and Suyash Shringarpure and Vinh Tran and Joyce Y. Tung and Xin Wang and Wei Wang and Catherine H. Weldon and Peter Wilton and Alejandro Hernandez and Corinna Wong and Christophe Toukam Tchakouté and Fredrik Ullén and Miriam A. Mosing},
  title = {{Using a polygenic score in a family design to understand genetic influences on musicality}},
  booktitle = {Scientific},
  year = {2022},
  doi = {10.1038/s41598-022-18703-w},
}
Figure