🧬 Genome-wide Study of Absolute Pitch Reveals Linkage to 8q24.21

📋 Study Overview

🎯 Research Question

Which regions of the human genome harbour variants that predispose to absolute pitch (AP), and do families of different ancestries share the same regions? The authors describe the work as “a first step toward identifying these genetic factors” — a search for chromosomal regions, not for a gene.

This was the first genome-wide linkage scan for AP. Before it, the genetic evidence consisted of familial aggregation and a sibling recurrence-risk ratio (λ_s between 7.8 and 15.1, controlling for early musical training). Because the families came from several ancestral backgrounds, the linkage analyses were run on three separate sample sets rather than on the 73 families as a whole.

The premise that motivated the study — not a conclusion from it

In the introduction the authors cite Saffran & Griepentrog (2001), who found that infants preferentially use absolute-pitch cues over relative-pitch cues in certain situations, “suggesting that all people might be born with AP but that the majority lose their AP abilities with age”. From that they formulate what they call “an attractive hypothesis”: that “genetic factors might extend this neurodevelopmental window to a duration sufficient to intersect with the onset of musical training”. No result in this paper tests that hypothesis, and the authors never connect it to any of the candidate genes.

Twin observations quoted in the introduction

Still in the introduction, and flagged by the authors as unpublished data (E.T., unpublished data) and explicitly “limited”, they mention that seven of ten monozygotic twin pairs were confirmed as concordant for AP, whereas nine of 20 dizygotic pairs were “confirmed or reported” to be concordant. This is background motivation, not a result of this study, and no heritability was estimated here.

🔬 Methodology

Family Recruitment and Assessment

Recruitment used the same online pitch-naming test and the same laboratory website already employed in Baharloo et al. (1998) and Athos et al. (2007); the paper does not say how many families came from those earlier cohorts. A family entered the linkage analysis only if AP ability was “documented in at least two family members” who were not simply a parent–child pair (multiplex design).

• 73 multiplex families with 281 individuals genotyped
• Ethnic composition:
  • 45 European families
  • 19 East Asian (E Asian) families
  • 8 Ashkenazi Jewish (AJ) families
  • 1 Indian family
• AP phenotype: an online pitch-naming test with a pre-defined threshold, plus a survey. A proband had to exceed that threshold on the test and report at least one relative with AP; the family then entered the analysis only if AP was documented by the test in at least two members who were not simply a parent–child pair. The paper does not restate the number of trials, the tolerance or the chance level — those are in Athos et al. (2007). AP is handled as a dichotomous trait, which the linkage analysis requires: every individual is either affected or not
• The three sample sets actually analysed (Table 1): Eu — 45 families, 184 genotyped, 108 with AP; Eu/AJ/I — 54 families, 220 genotyped, 128 with AP; E Asian — 19 families, 61 genotyped, 40 with AP. No analysis of all 73 families together is reported. AP parent–child pairs were not counted as affected relative pairs

Genome-wide Linkage Analysis

• 6,090 SNP markers spanning the entire genome
• Multipoint nonparametric linkage analysis (model-free, appropriate for a trait with unknown mode of inheritance)
• Stratified analysis by ethnicity to detect population-specific effects
• 10,000 gene-dropping simulations set the empirical thresholds separately for each sample set — in the Eu subset, LOD = 3.231 for significant linkage and LOD = 1.869 for suggestive linkage (LOD = 1.822 in the E Asian scan)
• The authors acknowledged the study was probably underpowered, especially for the E Asian and AJ subgroups given their smaller family counts

📊 Key Findings

1. The one significant result: 8q24.21, in the European subset

The single significant finding of the study came from the 45 European-ancestry (Eu) families analysed on their own, at marker rs3057. Table 2 of the paper reports every result together with its sample set, because the same marker behaves very differently depending on which families are included:

The LOD of 3.464 cleared the study’s own empirical significance threshold (LOD = 3.231, obtained from 10,000 gene-dropping simulations), with a genome-wide empirical p of 0.030 — which rises to 0.0364 once the Camp & Farnham correction for multiple analyses is applied. It is a significant result, but by a narrow margin and not yet replicated: the authors write that the study “bears extension by further recruitment within our own laboratory and replication by other groups”, that their LOD scores were “modest in comparison to the theoretical maximum” (34.31 for this subset), and that the study “was probably underpowered”. The abstract hedges in the same direction: “Though only one of these regions has yet reached statistical significance individually, we detected a larger number of independent linkage peaks than expected by chance overall”.

2. Three suggestive regions — and what their individual p-values look like

Three further regions in the Eu subset exceeded the empirical threshold for suggestive linkage (LOD = 1.869):

Taken one at a time, none of these is remarkable: a peak of that size would be expected by chance in 45% to 68% of genome scans. What carries statistical weight is the set of them — which is exactly why the paper needed the locus-counting analysis below.

8q21.11 is not an extension of the main signal. It lies on the same chromosome but roughly 53 cM away from 8q24.21 (86.732 cM against 139.741 cM), and by the study’s own criterion — peaks separated by 40 cM or more count as independent — the two are distinct signals. That independence is part of what supports the heterogeneity argument. It is also the region closest to the musical-aptitude peak (chromosome 8 at 92 cM) reported in Finnish families by Pulli et al. (2008), though the authors are careful: AP and musical aptitude may share linkage there, “though it was not the top linkage peak in either study”. They saw no evidence of AP linkage to 4q22, the major locus of that musical-aptitude scan, and none of their top peaks fall near the AVPR1A gene proposed for musical memory by Granot et al. (2007).

3. Evidence for locus heterogeneity: a locus count, not a story about families

4. The East Asian families: a null result, not a second locus

In the 19 East Asian families no peak reached even the suggestive threshold (LOD = 1.822), although regions on chromosomes 1, 3, 7, 13, 18 and 19 had peaks above LOD 1.0. In particular, “there was no evidence in the E Asian population for linkage in the region of significant linkage (8q24.21) from the Eu sample set”. The only overlap between the two scans was at 7q22.3. The eight Ashkenazi Jewish families likewise “do not show linkage to the top Eu linkage regions”.

These are null results, and the authors attribute part of them to insufficient power (19 and 8 families respectively). They are therefore compatible with genetic differences between populations without demonstrating them — and they are not evidence that some other locus produces AP in those families.

5. The four genes nearest the peak

In the UCSC genome browser the authors identify the four genes lying closest to the peak. The LOD−1 interval around the Eu peak spans 5.54 Mb, so the region holds a good many more genes than these four, and the paper does not prioritise any of them:

• GSDMC (gasdermin C)
• FAM49B
• ASAP1 (ArfGAP with SH3 domain, ankyrin repeat and PH domain 1) — “expressed in a variety of tissues, including the brain”
• ADCY8 (adenylyl cyclase 8) — “expressed almost exclusively in the brain” and “thought to play a role in learning and memory”

That is the whole of what the paper says about them. No sequencing was performed, no variant was identified, none of the four is singled out as more promising than the others, and the authors do not connect any of them to the critical-period hypothesis quoted above. The abstract opens by stating that “the molecular basis for AP remains unknown”.

💡 Main Conclusions and Implications

1. A locus found without a prior hypothesis — and it is a region, not a gene

No genome-wide linkage scan for AP had been attempted before, so 8q24.21 was located without any prior assumption about which genes to test. What was located, though, is a chromosomal region several megabases wide, in one subset of the sample. The authors are explicit that this is “a first step”: “AP is a complex trait, and the discovery of genes responsible for AP will provide the first step in unraveling the interplay between genetic predisposition and environmental influences.” They call for replication by other groups; the marker is a candidate for independent testing, not an established one.

2. More than one locus, within a population as well as between populations

The locus-counting result indicates that several loci can foster AP, and that this holds inside the European sample and not only across ancestries. The authors’ summary keeps two hedges that are worth reading in their own terms: the findings “provide strong evidence that at least one gene promotes the genesis of AP in individuals of European ancestry” and AP “probably results from genetic factors that vary both within and between different populations”. Note also that this is asserted “despite the fact that it is a dichotomous trait” — the dichotomy is a premise inherited from Athos et al. (2007), which the linkage method requires, not something this study tested.

3. Genes do not remove the environment

The paper’s own explanation for families in which AP does not appear in every carrier is incomplete penetrance, and it names a cause: “For some of our families, incomplete penetrance can be partially explained by a lack of early musical training in some family members”. The introduction is just as direct about training: “Musical training during a critical period of childhood development probably contributes to the acquisition of AP, but this training alone is insufficient; many people receive early musical training but do not develop AP”.

4. What this study does not address

Nothing in this design speaks to whether an adult can learn AP: there is no training, no behavioural task beyond the phenotyping, and no follow-up. Nor is any heritability estimated here. Readers looking for the adult-learning question should go to the training studies in this collection — Wong et al. (2025), Van Hedger et al. (2019) — not to this one.

5. Limitations acknowledged by the authors

• The study was “probably underpowered”, especially for the E Asian and AJ subgroups. For scale, the authors note that a study of 100 affected sibling pairs “could have greater than 90% power”, whereas “a similar study of only 40 affected sibling pairs would have 20%–70% power” — the Eu subset had 65 AP sibling pairs
• The LOD scores obtained were modest against the theoretical maxima for these samples: 40.03, 34.31 and 12.34 for the Eu/AJ/I, Eu and E Asian analyses respectively
• The markers were not fully informative — average polymorphism information content (PIC) = 0.35
• DNA could not be obtained from “informative relatives, such as parents”, in some families, “thus reducing power further”
• Incomplete penetrance, explained in part by a lack of early musical training in some family members
• The linkage interval is broad (5.54 Mb at LOD−1) and the causal variant or variants remain unknown; the authors suggest that linkage disequilibrium analysis “may help to narrow the interval”
• No sequencing and no functional validation of any candidate gene was performed
• The study “bears extension by further recruitment within our own laboratory and replication by other groups”

🎓 Study Strengths

• AP was established by an objective pitch-naming test with a pre-defined threshold, and had to be documented in at least two members before a family entered the analysis — the phenotyping came before the DNA collection
• Ethnically diverse sample, analysed as separate sets rather than pooled, which is what made the between-population comparison possible
• Genome-wide approach avoids the bias of candidate-gene studies
• 10,000 gene-dropping simulations set the significance thresholds empirically, for each sample set, instead of borrowing a conventional cutoff
• Family-based design reduces population stratification confounds
• Results are reported per sample set, with the non-significant and null analyses shown alongside the significant one

🔗 Related Research

• Same recruitment pipeline: Baharloo et al. (1998) — family study of AP from the same laboratory; Theusch used the same online pitch-naming test and website, but the paper does not say how many families came from that cohort
• Genetic component estimate: Gregersen et al. (1999) — prevalence and ethnic variation used to estimate the genetic component of AP; one of the sources of the λ_s = 7.8–15.1 cited by Theusch. (A different paper from the 1998 single-author editorial on this site, which is not indexed here)
• Early family study: Profita & Bidder (1988) — reported familial aggregation of AP; not a twin study, and later criticised as uncontrolled for background prevalence
• Tone language and AP: Deutsch et al. (2004) — tone-language speakers show elevated AP rates, suggesting gene–environment interaction
• Structural brain findings: Schlaug et al. (1995) — planum temporale asymmetry in AP musicians
• Where the collection disagrees: Levitin & Rogers (2005) argue that the search for a genetic component of AP “may be inherently doomed”, because genes cannot be separated from environment in a skill that has to be taught. Theusch et al. take the opposite side four years later. The site does not pick a winner between them

📖 Access Full Study

📚 Full Citation