Genome-wide association meta-analysis of human olfactory identification discovers sex-specific and sex-differential genetic variants

Abstract

Smelling is a human sense, expressing strong sexual dimorphisms. We aim to improve the knowledge of the genetics of human olfactory perception by performing an exploratory genome-wide association meta-analysis of up to 21,495 individuals of European ancestry. By sex-stratified and overall analysis of the identification of twelve odours and an identification score, we discovered ten independent loci, seven of them novel, with trait-wise genome-wide significance (p < 5 × 10^-8) involving five odours. Seven of these loci, including four novel ones, are also significant using a stricter study-wide significance threshold (p < 3.85 × 10^-9). Loci were predominantly located within clusters of olfactory receptors. Two loci were female-specific while one was sex-differential with respective candidate genes containing androgen response elements. Two-sample Mendelian randomization was applied to search for causal relationships between sex hormones, odour identification and neurodegenerative diseases. A causal negative effect was detected for Alzheimer's disease on the identification score. These findings deepen our understanding of the genetic basis of olfactory perception and its interaction with sex, prioritizing mechanisms for further molecular research.

Fig. 1. Combined Manhattan/Miami plots of traits with genome-wide significant variants showing overall and sex-stratified association results.

The x axis represents chromosomal positions of variants, and the y axis the negative log₁₀-transformed p values from two-sided linear regression tests of association. The genome-wide significance threshold (5 × 10⁻⁸) is shown as a red dashed line. A stricter significance threshold (3.85 × 10⁻⁹) accounting for the 13 investigated traits is shown as a red dotted line. Variants with genome-wide significance are coloured according to their associated trait. Index variants are annotated with a locus identifier and proposed candidate genes, if applicable. Annotations within boxes indicate novel loci. Bold labels indicate significant sex interactions (q_SIA < 0.05). Loci in parentheses are not considered independent. Labels are positioned according to the analysis group with the strongest association of the respective index variant. SNPs with −log₁₀(p) < 2 are not shown. SNPs with −log₁₀(p) > 20 are shown with a truncated value of 20.

Fig. 2. SNP-by-sex interaction analysis of index variants of the ten independent loci.

Displayed are estimates of effect size (β-coefficients of additive genetic effects) and their respective 95% confidence limits for males (x axis) and females (y axis). Three variants achieved significance after multiple testing correction (red = higher effect size in females, blue = higher effect size in males).

Fig. 3. Test for colocalization of male and female association signals at the 10 independent genome-wide significant loci.

We show the posterior probability (PP) for each of the five hypotheses. H0: no associations within the locus, H1: associations within males only, H2: associations within females only, H3: Association in both sexes but different causal variants, H4: association within both sexes with the same causal variant. Colour of the locus labels represents results of sex interaction analyses of index variants, namely red/bold: significantly higher effect in females, blue/bold: significantly higher effect in males, black/italics for loci with index variants not showing significant sex interactions.