Response to Comment on Large-scale GWAS reveals insights into the genetic architecture of same-sex sexual behavior – Science Magazine

Abstract

Hamer et al. argue that the variable ever versus never had a same-sex partner does not capture the complexity of human sexuality. We agree and said so in our paper. But Hamer et al. neglect to mention that we also reported follow-up analyses showing substantial overlap of the genetic influences on our main variable and on more nuanced measures of sexual behavior, attraction, and identity.

Genetic research on sexuality had been constrained by the unavailability of samples large enough to achieve the statistical power required to detect variants with small effect sizes typical of complex traits. To address this, our study aimed to maximize statistical power in two ways: (i) combining the largest samples available with sexuality data from UK Biobank and 23andMe, and (ii) choosing the variable that yielded the largest effective sample size. Only the dichotomous variable we used in the main genome-wide association study (GWAS) satisfied both considerations. This variable we chose is straightforward and easy to understand, and was clearly described in the paper, including characterizing it as ever versus never had a same-sex partner as suggested by Hamer et al. (1). We made it clear that this variable did not capture the diversity and complexity of sexual behavior, and we did not intend or claim to measure sexual orientation or attraction with this variable. However, we reported follow-up analyses with subsets of the data and independent samples that showed overlap in the genetic signal for our main variable and for traditional measures of sexual orientation based on sexual attraction and identitycontrary to Hamer et al.s assertion that the study did not in fact investigate attraction or sexual orientation.

As is common for large-scale genetic studies, our analyses were constrained by the available phenotypic data, which were not collected for the particular purposes of our study. The UK Biobank, which comprised most of the available data, did not include sexual attraction or identity items. The only UK Biobank items relevant to sexual behavior were one question asking whether the participant had ever had a same-sex partner, and two questions asking their lifetime number of opposite-sex and same-sex partners (9% fewer respondents). Only the variable based on the dichotomous item maximized the absolute sample size. This variable also had the advantage of having an equivalent variable in the 23andMe sample (i.e., Other sex only response versus other responses to the question With whom have you actually had sex?). It also had a far greater effective sample size than the other dichotomous variables with a direct equivalent in 23andMe, namely restricting the non-heterosexual group to those who had only had same-sex partners, which in the UK Biobank can be derived from the items on numbers of same- and opposite-sex partners (effective UK Biobank sample size N = 53,688 for our main variable, versus N = 9775 when comparing heterosexuals with those who only had same-sex partners; effective sample size derived from the formula 4/[(1/Ngroup 1) + (1/Ngroup 2)].)

Without nuanced measures of behavior, fantasy, attraction, and identity in the large majority of our sample, our best option was to use the variable with maximal power available in the full sample (while acknowledging its limitations) and perform more nuanced follow-up analyses in subsets of the data to explore these important research questions about the complexity of sexuality. Hamer et al. seemingly disregard these follow-up analyses, which show evidence for substantial overlap in the genetic influences on sexual behavior with the kinds of measures that Hamer et al. recommend. First, in the UK Biobank there was a genetic correlation between our main dichotomous variable and a continuous variable measuring the proportion of total partners who were same-sex partners [rg = 0.92; supplementary materials of (2), p. 12]. Second, the 23andMe data included Klein scales for sexual fantasy, identity, attraction, and behavior, and these measures were genetically correlated with the main dichotomous measure in the UK Biobank [rg = 0.83, 0.79, 0.75, and 0.70, respectively; table S5 of (2)]. Third, we replicated three single-nucleotide polymorphisms identified by the main GWAS in a much smaller independent sample [MGSOSO; table S10 of (2)] whose comparison groups were predominantly heterosexual and predominantly homosexual individuals, per self-reported identity and feelings. Fourth, polygenic scores based on the main GWAS significantly predicted the identity-based groups in MGSOSO [table S12 of (2)] and continuous measures of same- versus opposite-sex attraction in two other small independent samples of young adults [Add Health and CATSS; tables S13 and S14 of (2)]. So, again, although we did not intend or claim to measure sexual identity or attraction with our main dichotomous variable, follow-up analyses showed that the genetic signal does substantially overlap for these phenotypes.

Responding to our suggestion that some of our findings cast doubt on popular measures of sexuality, Hamer et al. contend that genetic research cannot inform sexology research and that such an idea is an inversion of the scientific process. We disagree with this contention. To suggest that either sexology research or genetic research has supremacy in scientific enquiry is misguided. Both sexology research and genetic research can and should inform each other. Our genetic analyses revealed insights into the underlying structure of variation in sexual behavior that could not have been obtained using traditional methods of sexology researchfor example, that there is partial overlap in the influences on male and female sexual behavior.

The traditional and most popular measure of sexual orientation is the Kinsey scale (Fig. 1), which is bipolar and implies a continuum between exclusive heterosexuality and exclusive homosexuality, measuring the relative incidence of a composite of same-sex versus opposite-sex sexual behavior and psychological responses (not simply preference, as claimed by Hamer et al.). A concern with the Kinsey scale is that it inappropriately measures homosexuality and heterosexuality on a single dimensional scale, making one trade-off of the other (3). This enforced trade-off would not be a problem if it reflected the true underlying structure of individual variation in sexuality, such that the amount of same-sex behavior (and/or psychological responses) were indeed perfectly inverse to their opposite-sex counterparts. However, other research suggests that this is not the case (46). Individuals can be high on both same-sex and opposite-sex behavior or attraction (some bisexual individuals), and individuals can be low on both (asexual). This variation is not captured by a bipolar scale. Our findings (2) reinforced the point at the genetic level: The genetic variants that increase the likelihood of having had any (versus no) same-sex partners do not increase the likelihood of having a greater (versus lesser) proportion of total partners who are same-sex partners. That is, on the genetic level, there is no one continuum from exclusively opposite-sex to exclusively same-sex behavior. The 23andMe data showed genetic correlations (rg = 0.83 to 1.0) of same-sex behavior with attraction and fantasy [figure S7 of (2)], which suggests that the finding on same-sex sexual behavior might extend to these other aspects of sexuality too. For future genetic research on sexuality, if the phenotypic measure does not reflect the structure of the underlying genetic influences, then the precision and accuracy of the findings will be impaired. For this reason, we, like others [e.g., (35, 79)], suggest that sexual attraction, behavior, and feelings toward men and women be measured separately in future research.

The ratings are based on both psychologic reactions and overt experience. [Source: figure 161 of (10)]

Although we would much prefer even massive biobank-based samples to have deep phenotyping on our topic of interest (i.e., sexuality measures), data of this nature are not currently available. Generally, there is a practical trade-off between phenotypic detail and sample size. Our approach was to acknowledge the limitations but make reasoned use of the available data to move the field forward, recognizing that others may prefer to avoid examining such datasets altogether. In any event, we share Hamer et al.s concern to minimize public confusion, which is why we liaised with community groups on reporting and communication strategy, provided lay-oriented explanatory boxes 1 and 2 in our manuscript, developed a companion educational website (www.geneticsexbehavior.info), welcomed the public posting of alternative perspectives and concerns about our work (www.broadinstitute.org/news/perspectives-complex-genetics-same-sex-sexual-behavior), and held an informational press conference.

R. L. Sell, in The Health of Sexual Minorities: Public Health Perspectives on Lesbian, Gay, Bisexual and Transgender Populations, I. H. Meyer, M. E. Northridge, Eds. (Springer, 2007), pp. 355374.

A. C. Kinsey, W. B. Pomeroy, C. E. Martin, Sexual Behavior in the Human Male (Saunders, 1948).

See the article here:
Response to Comment on Large-scale GWAS reveals insights into the genetic architecture of same-sex sexual behavior - Science Magazine