When we move in time with a musical beat, we are harnessing one of the (arguably) most exquisite feats of perception and coordination of everyday human existence. As knowledge has evolved over the past few decades about the behavioral and neural mechanisms that support human rhythm skills, scientists’ curiosity about how rhythm traits evolved in humans and whether the feat of moving in synchrony with a musical beat has a signature in the human genome.
In a new study published this week in Nature Human Behaviour, we investigate the genetic architecture of beat synchronization (i.e., moving in time with a musical beat) using advanced genomic analysis methods. For this study, we assembled an international multi-disciplinary team of experts in complex trait genetics, music cognition, evolutionary biology, evolution, musicology, and neuroscience, and collaborated with personal genetics company, 23andMe, Inc. With the formidable sample size made possible by their research participant pool, N=606,825 gave their informed consent to participate and answered the question “Can you clap in time with a musical beat?”. We conducted a GWAS on beat synchronization by comparing the frequencies of genetic alleles at over 8 million regions of the genome in individuals who responded Yes (~92%) versus No (~8%) to this question. We were particularly careful about the ethical and societal implications of our study: notably we paid significant attention to the ways that the results should and should not be interpreted (discussed later in the blog).
Can we reliably assess beat synchronization with a single questionnaire item?
In short, yes! The clap-beat self-report is a good proxy for beat synchronization. In a series of internet-based experiments (run in separate samples from the genetic studies), we performed phenotype validations (experiments validating that the self-report questionnaire item) by asking people to complete musical rhythm tasks and respond to questionnaires.
Results showed associations between self-report to the clap-beat item and accuracy on these tasks. Here we used REPP, a novel technology to measure tapping responses via internet, in real time. As participants listened to music, we recorded their tapping responses using the computer microphone and extracted exactly when they tapped relative to the musical beat. The “objective” (measured) degree of synchronization with music in the tapping experiment is highly correlated with subjective (self-) reports of beat synchronization. These associations held even after controlling for important demographic variables. Together, the phenotype validation experiments led us to conclude that the self-report item is a suitable proxy for beat synchronization. When participant time and technology are limited, such as when working in large-scale epidemiological efforts where multiple questions are asked to monitor other health-related goals, the value of a single item is particularly significant.
Broadly characterizing the genetic architecture of beat synchronization
The GWAS revealed 69 independent genetic loci associated with beat synchronization, thus confirming that the genetic architecture of this trait is highly polygenic. Polygenic traits (also called “complex traits”, i.e., that do not follow a simple Mendelian pattern of inheritance) involve cumulative (small) contributions from many, many different genetic loci. This polygenicity is illustrated in the Manhattan plot of the GWAS below. The SNP-based heritability (i.e., total phenotypic variability accounted for by genetic variation) was estimated at 13-16% on the liability scale, in line with other complex traits in the cognitive/neurological domain. In the paper we discuss in more detail a few loci that are particularly interesting, but it is very important to keep in mind that each locus contributes only a small amount of phenotypic variation (the variability in responses to the self-reported questionnaire) and thus there may be more value in examining the results collectively.
The results of this first large-scale GWAS of a musicality trait are significant in pinpointing genetic alleles that cumulatively form a robust association with inter-individual variability in the beat synchronization trait. With the resulting GWAS summary statistics, future avenues of research are now also open to understanding the intricacies of the genetic architecture of this trait, beyond even what we have begun to show in the current work. For instance, our ongoing and future work is investigating the relationship between beat synchronization genetics and other aspects of musicality such as music aptitude (leveraging twin data: Wesseldijk et al., in revision) and music engagement (Gustavson et al., under review). We are also very interested in looking at pleiotropy (shared genetic architecture) between beat synchronization and communication skills, both in humans (i.e. speech and language skills Nayak et al., 2022, PsyArXiv), and with vocal learning in songbirds (Gordon et al., 2021).
So did this study discover “a rhythm gene”? Or are there many rhythm genes? Do these genes influence other traits as well, or are they restricted to rhythm?
These results demonstrate that there are many, many genes that influence beat synchronization. We identified alleles (specific genetic variants on those genes) that explain a portion of population's variability in the beat synchronization aspect of rhythm skills, but there is likely no such thing as a single “rhythm gene” that has a deterministic effect. With such a large set of results, we used a number of methods to investigate biological and health functions of the variants and genes found to be linked to rhythm traits, described in the next sections of the blog.
What do these genetic associations reflect, biologically? Can we begin to bridge genetic variation to the neurobiology of beat synchronization?
We conducted a series of computational enrichment analyses to understand the functions of the SNPs and genes associated with beat synchronization. Notably, we found enrichment (greater than chance associations) of genes linked to many facets of central nervous system function, including gene expression in auditory and motor regions of the brain previously linked to beat perception and synchronization. We also found enrichment of brain-tissue-specific enhancers (i.e., SNPs that play a key role in gene regulation), in fetal and adult brain tissue. Therefore, these findings are an initial bridge between the genetics and neural basis of rhythm processing, suggesting that genetic variation influences development and function of specific brain circuitry underlying beat synchronization. With this crucial groundwork, future work will focus on examining the relationships between genetic variation and specific neural endophenotypes underlying beat synchronization at a more granular level.
Is there evidence of evolutionary signatures of rhythm?
Evolutionary enrichment analyses produced limited evidence for enrichment of beat synchronization among human-accelerated regions (i.e., that have experienced significant human-specific shifts in evolutionary pressure), among them a SNP on GBE1, a gene previously linked to cognitive and neuromuscular traits.
How do we know the polygenic model is valid, i.e. does it truly reflect some aspect of the biology of musicality?
For this analysis we want to transport the reader to Nashville, TN, otherwise known as Music City, USA, home of Vanderbilt University Medical Center (primary affiliation of many of the study authors). In a prior study, Niarchou and collaborators had algorithmically identified musically active patients in the Vanderbilt University Medical Center’s health care research database. Here, we validated the polygenic architecture of the GWAS using a Polygenic Scoring (PGS) method, where we derive weights from the GWAS and then apply them to an independent sample to predict a related trait. In other words, in PGS analysis we create a score with the joint contributions (weights) of the alleles that predicts our musicality phenotype (here, a beat synchronization PGS). In this case, we applied PGS weights to the genotypes of ~1200 musically active patients in Vanderbilt’s BioVU database from the prior study, and compared to an additional ~4900 individuals matched on demographics and thought to represent the broader population (i.e. not necessarily musicians). The musically active group on average had higher PGS than the control group, demonstrating that the genetic architecture of beat synchronization, as measured from the 23andMe sample, is indeed connected to musicality in a broader context.
Does beat synchronization share genetic architecture with other traits?
We explored shared genetic architecture (pleiotropy) with other complex traits of interest using the genetic correlation method (part of the LDScore regression software), which allows for the assessment of genetic overlap from separate GWASs. A highlight of the results was that the genetic architecture of beat synchronization (an aspect of musical rhythm) was linked to the genetics of multiple biological rhythms including breathing, walking pace, and circadian chronotype (evening versus morning person). We further explored shared genetic architecture between beat synchronization and a constellation of interrelated traits: motor function, lung function, and processing speed using the Genomic SEM method. The genetic signal of a common factor on these rhythm-related traits was enriched for gene expression in cerebellum, which is known from the neuroscientific literature to play a role in various timing functions (including beat processing, time perception, breathing, speech, postural control, and motor coordination). Taken together, results of these analyses may have implications for physical and cognitive function during aging.
We then followed up on the genetic correlations by asking a separate group of participants, in a new behavioral sample, subjective report questions that we expected to be correlated with the ones that we found in the GWAS study. We found that indeed, more accurate tapping synchronization task performance was associated with self-reported evening chronotype and better breathing function. Therefore, phenotypic results mirror and replicate genetic correlation results.
And what are the limitations of the study that can be addressed in future research?
The primary limitations of the study are the simplicity of the phenotype in the GWAS, and the constraining of the study sample to only European ancestry; these limitations can be addressed in future research by using more nuanced phenotypes (for example, questionnaires about musicality that utilize a Likert scale for responses, or directly measuring musical skills in the GWAS participants) and conducting additional GWAS’s in other genetic ancestries (genetic ancestry has to be handled correctly due to effects of population stratification that can confound GWAS results). We hope and plan to pursue such work in the near future, focusing on diverse populations!
Possible future directions can also involve more sophisticated behavioural paradigms and larger global sample. For example, musical rhythm has been a focus of recent cross-cultural study that collected a nuanced and complex measure of rhythm representations structure, far and beyond a single question single report, in 39 groups from 15 countries around the world.
How can we apply socially and ethically responsible principles of research conduct to these lines of research?
Individual differences, human genetics, and musicality each have a long history of being discussed in contexts that are embedded in racism (including the abhorrent and scientifically inaccurate theories of eugenics). It is especially important to be aware of this history, and proactively work towards principles of diversity, equity, and inclusion in this research, moving forward. To this end, several members of the authorship team joined forces with two outside experts in musicology and bioethics to put forth a set of guidelines for ethically and socially responsible conduct of musicality genetics research. These guidelines are discussed in a recent preprint, and also summarized in the Nature Human Behaviour paper (see Box 1). In brief, incorporating principles of ethically and socially responsible conduct of musicality genetics research can be infused into each stage of the research lifecycle: study design, study implementation, potential applications, and communication about the research.
For instance, it is crucial to understand that genome-wide associations with beat synchronization are not deterministic and that we cannot make deterministic individual inferences or rankings based solely on genetics. Environment also plays a major (and not yet well-understood) role in influencing individual rhythm skills! It is important to note that biology (including genetic variation) can draw us into different cultural experiences, such that biological processes and human culture are not independent from one other.
Moreover, we unambiguously welcome the diversity of musical traditions, competencies, and diverse experiences of music across cultures; this diversity and the scientific discoveries on the horizon are strengthened by the continued disavowal of research founded on eugenic and ethnocentric principles. The authors of both the GWAS article and the Commentary urge the research community and broader public to understand the broader social (and scientific) contexts of this work, and to take a firm stance on only using the knowledge resulting from this study for positive outcomes (i.e., improving societal well-being and health).
Full Citation information.
Niarchou, M., Gustavson, D.E., Sathirapongsasuti, J.F., Anglada-Tort, M., Eising, E., Bell, E., McArthur, E., Straub, P., 23andMe Research Team, McAuley, J.D., Capra, J.A., Ullén, F.U., Creanza, N., Mosing, M.A., Hinds, D.A., Davis, L.K.*, Jacoby, N.*, & Gordon, R.L.* (2022). Genome-wide association study of musical beat synchronization demonstrates high polygenicity. Nature Human Behaviour. https://doi.org/10.1038/s41562-022-01359-x *Indicates co-senior authors
To support the scientific community and general public in understanding the nuance of our study results, we have put together a study FAQ, updated in real time. https://www.vumc.org/music-cognition-lab/FAQbeatGWAS
Please contact Dr. Reyna Gordon (firstname.lastname@example.org) for more information about the study.
How was this work funded?
This work was supported in part by an NIH Director’s New Innovator award #DP2HD098859. Complete list of funding sources are reported in the article.