Women choose romantic partners resembling their father in body odour

Women choose romantic partners resembling their father in body odour | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Social Sciences - Article Women choose romantic partners resembling their father in body odour Jan Havlíček, Lucie Jelínková, Zuzana Štěrbová, Robert Hanus, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-2531352/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Women tend to choose partners who resemble their father in certain characteristics. In non-human mammals, similar parental imprinting-like effects are often odour-mediated1,2, which requires not only learning of parental odour but also similarity between parents and prospective mates in the microbial communities responsible for production of the critical volatile compounds. Here, in light of growing recognition of human capacity for social olfaction3,4, we tested the possibility that women’s preferences are shaped by paternal phenotype. Axillary body odour samples from women’s fathers and romantic male partners were evaluated for similarity at the perceptual level as well as in both their microbiome diversity (using 16S rDNA sequencing) and chemical composition (using GC×GC-TOFMS). We found that partners’ body odour was perceived as more similar to paternal odour than would be expected by chance. In parallel, similarities in both microbiota composition and chemical profiles were significantly higher in partner-father dyads than in randomly generated dyads. While women’s relationship quality with the father did not predict father-partner similarity in microbial or chemical profiles, it was positively associated with perceived odour similarity. Our results thus reveal a suite of mechanisms by which parental odour cues can serve as a template in human mate choice. Biological sciences/Psychology/Human behaviour Biological sciences/Microbiology/Microbial communities/Microbiome Biological sciences/Chemical biology/Chemical ecology Figures Figure 1 Figure 2 Figure 3 Introduction Cross-fostering studies in mammals show that learnt olfactory cues in the postnatal environment affect subsequent mate choice in adulthood, leading to phenotypic similarity between adoptive parents and the partners of their adoptive offspring 2,5 . What makes this observation even more remarkable, compared to visual or acoustic forms of phenotype-matching, is that such odour-mediated behaviour is not an entirely autonomous process, but appears to be critically dependent on the commensal microflora. Thus, bacterial action on less volatile molecules produced by the mammalian host is required to generate the volatile metabolites which form the host’s odour cues 6–8 . If this is the case, then we would hypothesise that in a species where parental odour characteristics influence mate preferences, we should observe similarities not only in how parental and mate odours are perceived, but also in the microbial communities that are responsible for the odour’s generation and the resulting chemical profiles. Here we test this hypothesis for the first time, addressing human mate preferences which appear to form during childhood and where parental characteristics play a pivotal role 9 . For example, women tend to choose partners resembling their father in various physical 10,11 and personality 12 characteristics. Furthermore, levels of facial similarity are positively modulated by the quality of relationship with the father during childhood 13,14 . Some evidence also suggests that women prefer odours of men who have HLA alleles in common with their father 15 , but whether this affects actual partner choice or leads to odour similarity between women’s partners and their fathers has not yet been addressed. To answer such questions, we investigated similarity between fathers and the romantic partners of their daughters. For the first time in any species, we did this at each of the three main steps in the hypothesised pathway, examining i) perceptual ratings of odour similarity, ii) similarity in their microbiomes, and iii) similarities in their chemical profiles. Perceived similarity To test perceived similarity in body odour we recruited 67 male dyads (a woman’s long-term romantic partner and her father) (Fig. 1a). All women grew up in the same household as their father at least up to 8 years of age. The participants refrained from using personal cosmetics, eating aromatic food and from other activities affecting body odour for 2 days preceding the sampling and subsequently collected their axillary body odour samples on cotton pads worn for 12h (overnight) (Fig. 1b). Perceived similarity between the target father sample and the partner was tested by a 4-choice match-to-sample test 14 (Fig. 1c), in which female panellists (unacquainted with the odour donors) were asked to smell and then order the 4 alternatives (the partner and three age-matched distractors) based on their similarity to the target father sample. Each panellist assessed 5—6 sets of stimuli and each sample was assessed by at least 20 independent panellists (N = 306) which makes 1566 match-to-sample tests in total. Panellists selected the partner’s body odour as the most similar to the father’s body odour in 35.5 % of cases, which is significantly more frequent than chance (25 %, Chi-square = 92.84, p < 0.001). Furthermore, the frequency of selection in the second-, third- or least-similar position decreased in a near-linear fashion (Fig. 1d). To analyse all these selections while accounting for non-independence of individual ratings and target’s identity, we used an ordinal probit mixed model. This analysis revealed substantial differences in similarity ratings between partner and non-partner samples (estimate [± 95% posterior credibility intervals] = -0.48 ± [-0.64 ~ -0.32]). According to the parameter estimates, partner samples had a much greater chance of being rated as most similar or second most similar to the target father sample than the non-partner samples. Conversely, non-partner samples were more likely to be classified as the least similar or the second least similar. Microbiome similarity The axillary microbiota was analysed by amplicon sequencing of the 16S rRNA gene. The resulting microbiota profiles revealed a dominance of bacteria belonging to the genera Staphylococcus (42.2% of sequences), Corynebacterium (31.1% of sequences) and Anaerococcus (10.0% of sequences, Fig. 2a). Microbial alpha diversity, composition, and predicted microbial functions showed only minor differences between the fathers and the partners (Extended Data Fig. 1a, Extended Data Table 2a and Extended Fig. 2a-c) and there were no systematic differences between the microbiota profiles of the left and right axillae. Moreover, the left and right microbiota were closely correlated at the individual level, both in terms of alpha diversity (Pearson correlation: r = 0.796, p < 0.0001, for microbiota richness and 0.797 for Shannon diversity) and composition (Extended Data Fig. 1c,d). Therefore, for all subsequent analyses, we used individual-level microbiota profiles generated by aggregating the left and right microbiota for each individual. To test if there was a higher within-dyad similarity than expected by chance, we compared average microbiota dissimilarity between actual father-partner dyads with a distribution of average dissimilarities for randomly paired fathers and partners (n = 999 permutations). This approach provided consistently significant results for dissimilarities calculated based on amplicon sequence variant (ASV) prevalence and relative abundances. These results remained unchanged whether we analysed the whole dataset or only a subset of pairs with no missing sample (i.e., profiles of both left and right axilla of both father and partner, n = 49 dyads, Extended Data Table 2b) and were also significant if we analysed similarity in predicted microbiota functions (Extended Data Fig. 2c,d). To assess the potential impact of environmental bacteria on father-partner axillary microbiota similarity, environmental microbiota samples were also collected. The taxonomic content of environmental microbiota samples was clearly distinct from axillary microbiota (Extended Data Fig. 3a,b). As in the case of axillary microbiota, environmental samples of father-partner dyads exhibited higher similarity than randomly paired individuals (p = 0.019, SES = 2.24 for Bray-Curtis and p = 0.018, SES = 2.47 for Jaccard dissimilarities). According to differential abundance analysis, nine bacterial ASVs were overrepresented in the environmental samples (Extended Data Fig. 3c), representing 39.5% of high-quality sequences of control samples but only 1.8% of reads of the axillary microbiota profiles. Importantly, when these ASVs were excluded from the axillary microbiota profiles, we still detected higher similarity of axillary profiles than expected by chance (p = 0.008, SES = 2.53 for Bray-Curtis and p = 0.004, SES = 2.94 for Jaccard dissimilarities) (Extended Data Fig. 3d). However, similarity between partners’ and fathers’ environmental samples did not deviate from random expectation after the exclusion of ASVs overrepresented in environmental microbiota (p = 0.319, SES = 0.45 for Bray-Curtis and p = 0.179, SES = 0.91 for Jaccard dissimilarities) (Extended Data Fig. 3e). In conclusion, these analyses suggest that similarity in axillary microbiota of fathers and partners is not caused by the similarity in environmental bacterial pools. Chemical compounds similarity In 41 father-partner dyads, we succeeded in analysing, in a single measurement sequence, the left and right axillary odour samples using comprehensive two-dimensional gas chromatography coupled with time-of-flight mass spectrometric detection. Even a visual inspection of the obtained chromatograms revealed, in some cases, a marked similarity of the chemical profiles between the father and the corresponding partner, as shown on selected chromatograms in Fig. 3a. The chromatograms were then processed using the recently developed tile-based alignment algorithm in the program ChromaTOF-Tile. Retrieval of more than 1,500 chemical features initially detected across all samples according to the inclusion criteria inferred from the pilot experiment (F-ratio = 2, S/N ratio = 50) followed by manual removal of artifact peaks resulted in the final set of 341 consistently occurring human sample-specific compounds. Their tentative identifications and peak volumes are listed in Supplementary Table Dataset 2. Among these compounds, we identified an expected set of human body odour analytes, especially homologous series of fatty acids, wax esters, hydrocarbons, aldehydes, ketones, amides, and steroids. Relative abundances of square root-transformed peak volumes of the 341 compounds were treated as three separate datasets to decipher the chemical similarities among all participants for left, right and left-right averaged axillary samples. Fig. 3b shows a representation of the factor scores of the first three components of a principal component analysis (PCA) for left-right averaged samples. The distributions of father and partner scores largely overlap and do not indicate any consistent age-related separation between the two groups. The factor scores were then used to calculate the Euclidean distances among all pairs of samples, a proxy for chemical dissimilarity. These distances are plotted for fathers and corresponding partners in Fig. 3c. Comparison of the averaged within-dyad father-partner distance with the distribution of average distances for randomly paired fathers and partners (n = 999 permutations) revealed that the observed distances are significantly shorter than would be expected by chance (p = 0.001; Fig. 3d). Likewise, within-dyad father-partner distances were significantly shorter than the average distances of the corresponding fathers to all extra-pair partners (Wilcoxon Signed Rank Test, p = 0.0006; Fig. 3e). We obtained similar results when the same calculations were performed separately for left and right axillary samples (Extended Data Fig. 4). That is, average father-partner distance was significantly shorter than would be observed by chance (n = 999 permutations, p = 0.001 for both right and left axillary samples), and within-dyad father-partner distances were significantly shorter than average distances between individual fathers and all extra-pair partners (Wilcoxon Signed Rank Test, p = 0.006 for left and p < 0.0001 for right axillary samples). At the same time, the distances between all possible pairs of participants, including the observed father-partner distances, were significantly correlated between the right and the left axillary samples dataset (linear regression, r 2 = 0.473, p < 0.0001 for all pairs, r 2 = 0.391, p < 0.0001 for within-dyad father-partner distances; Extended Data Fig. 4f), as were also the distances between the left and the right axillary samples dataset, and the left-right averaged dataset (Extended Data Fig. 4i,j,k, respectively). The factor scores and variable loadings for the first three principal components and the resulting Euclidean distances between all possible pairs of participants are listed for the three analysed datasets in Supplementary Table Dataset 2. Relationship quality with father Previous studies indicate a positive association between reported quality of the relationship between daughter and father and the phenotypic similarity between her romantic partner and her father. Therefore, we also examined the perceived father-partner odour similarity and three measures of relationship quality. The first two measures concerned the daughter’s relationship quality with her father during childhood (assessed using the s-EMBU scale and by a single-item overall assessment of the relationship quality during childhood), while the third addressed the quality of their current relationship (using the ECR-RS scale). Correlations between these variables were generally moderate (Spearman's rho < 0.4; Supplementary Table3). We found that perceived similarity was statistically significantly associated with the overall childhood relationship quality measure (Spearman’s rho = -0.28, p = 0.02), but not with any of the s-EMBU and ECR-RS subscales (Fig. 1e). None of the relationship quality measures showed significant association with either father-partner microbiome similarity (Extended Data Fig. 6) or chemical profile similarity (Extended Data Fig. 7). Discussion Our results demonstrate the existence of an olfactory parental imprinting-like effect in humans which is consistently observed across connected but different levels of analysis. First, we show that the axillary odours of a woman’s father and romantic partner are perceptually more similar than expected by chance. Second, since axillary odour results from microbial activity at the skin surface, we also compared the composition of axillary microbiota in the same father-partners dyads. Using microbiota profiling based on 16S rRNA amplicon sequencing, we discovered significantly higher similarity in microbial diversity in these actual dyads than in randomly generated permutations. Third, by analysing the chemical profiles underlying the fathers’ and partners’ odours using GC×GC-TOFMS, we found significantly higher similarity in father-partner dyads as compared to randomly generated pairs. Parental imprinting effects on adult mate choice have been demonstrated in cross-fostering studies in various mammals, from rodents 1 to ungulates 5 . While the nature of human research means that similar studies are correlational, similarity between a romantic partner and the opposite-sex parent has been reported in several other appearance-based characteristics, such as facial shape 13 or hair and eye colour 10 . The formation of such preferences may occur either continuously during childhood (perhaps, until adulthood) or be restricted to a specific sensitive juvenile period. In non-human mammals, olfaction-mediated social preferences form relatively early in ontogeny 16 but might be modulated by subsequent experience 17 . In humans, this remains poorly understood 12 . Earlier studies suggested that 1-5 years might be the most sensitive period 19,20 , while more recent investigation points to the early teenage years 21 . We suggest that similarity between a woman’s father and romantic partner results from a process of social learning of parental characteristics during childhood. A major assumption of this is that such characteristics are stable over a long period of time (i.e., between one’s childhood and young adulthood), as is known to occur in human faces 22,23 . While no similar data is available for body odours, studies on individual odour recognition 24 and genetic influences 25 do suggest such stability. An alternative explanation for the reported similarity between parent and romantic partner is that it arises as a by-product of self-similarity preferences 26 . Body odour of parents and children 27 or siblings 25 are perceived as more similar compared to non-related individuals. While we cannot rule this out, previous studies indicate that parental effects are stronger than self-similarity preferences 28 . A further alternative is that perceived odour similarity of fathers and partners, and the underpinning similarity in chemical profiles that we report here, arises not due to the hypothesised imprinting-like effects, but rather as a result of the daughter’s association with the two men. For example, learned dietary preferences and lifestyle characteristics in the natal household could be replicated in the woman’s adult home that she shares with the partner. If these behavioural patterns shape the microbial populations present in the natal and adult homes, they could in turn generate similarities in the chemical profiles of those living there, and of course in their emergent odour. However, analysis of the microbiome in environmental samples collected at the homes of the participants, demonstrated that the reported pattern cannot be explained simply by similarities in their environment. Theoretical models predict that mate preferences should be flexible to current environmental fluctuations 29 . Why, then, would selection shape adult mate preferences according to parental cues so early in ontogeny? The answer may be that complete flexibility is demanding to maintain or potentially disadvantageous in a stable environment. Using parental characteristics as a template for own mate choice might be a mechanism for outbreeding avoidance or preference for time-tested and locally adapted phenotypes . Finally, some studies have indicated positive associations between the level of similarity between the opposite-sex parent and the offspring’s romantic partner and the quality of the relationship between parent and offspring 13,14 . Although its robustness should be confirmed in future studies, our results partly support such an association between perceived odour similarity and relationship quality. The fact that relationship quality was, in contrast, not associated with either microbial or chemical profiles, suggests that the learned odour which appears to influence mate preference is formed from a subset of the entire chemical profile. This is to be expected given the chemical diversity of axillary odour and the multiple forms of social information it appears to carry 30 . Altogether, our findings reveal a chain of mechanisms that influence human social odour preferences, and open new directions in how early family influences may impact on social functioning in adulthood and in romantic relationships in particular. Declarations Data availability The data that support the findings of the current study are available at Supplementary Table Dataset 1 and Dataset 2. Acknowledgements The study was supported by a Czech Science Foundation grant (no.18-15168S). R.H., R.B. and P.K. were supported by the Institute of Organic Chemistry and Biochemistry, Czech Academy of Sciences (RVO 61388963). We are indebted to Kristýna Dachsová, Anna Fišerová, Kristýna Molnárová, Žaneta Pátková, Kateřina Roberts, and Kristýna Šípková for their help with data collection, Vladimír Kunc for programming the rating application and technical support, Pavel Šebesta and Vít Třebický for their help with the manuals, and to all the participants for their cooperation. Author contributions Conceptualization: L.J., Z.Š., R.H., J.K., and J.H.; methodology: L.J., Z.Š. R.H., J.K., P.K., J.T.F., and J.H.; project administration: L.J., and Z.Š.; data collection: L.J., Z.Š.,D.S., and J.T.F.; chemical analysis: R.H., P.B., and R.B.; microbiome analysis: J.K., and L.S.; statistical analysis: L.J., R.H., and J.K.; visualization: L.J., R.H., J.K., and D.S.; funding acquisition: J.H.; writing of the original draft: L.J., R.H., J.K., and J.H.; writing, review and editing: L.J., Z.Š., R.H., J.K., P.K., R.B., L.S., J.T.F., D.S., S.C.R., and J.H. Competing interests The authors declare no competing interests. Reprints and permissions information is available at www.nature.com/reprints. References Penn, D. & Potts, W. MHC-disassortative mating preferences reversed by cross-fostering. Proc. R. Soc. London Ser. B-Biological Sci. 265 , 1299–1306 (1998). Beauchamp, G. K. & Wellington, J. L. Cross-species rearing influences urine preferences in wild guinea pigs. Physiol. Behav. 26 , 1121–1124 (1981). McGann, P. John. Poor human olfaction is a 19th-century myth. Science (80-. ). 356 , 6338 (2017). Ravreby, I., Snitz, K. & Sobel, N. There is chemistry in social chemistry. Sci. Adv. 8 , (2022). Kendrick, K. M., Hinton, M. R., Atkins, K., Haupt, M. A. & Skinner, J. D. Mothers determine sexual preferences. Nature 395 , 229–230 (1998). Theis, K. R. et al. Symbiotic bacteria appear to mediate hyena social odors. Proc. Natl. Acad. Sci. U. S. A. 110 , 19832–19837 (2013). Zomer, S. et al. Consensus multivariate methods in gas chromatography mass spectrometry and denaturing gradient gel electrophoresis: MHC-congenic and other strains of mice can be classified according to the profiles of volatiles and microflora in their scent-marks. Analyst 134 , 114–123 (2009). Archie, E. A. & Theis, K. R. Animal behaviour meets microbial ecology. Anim. Behav. 82 , 425–436 (2011). Boothroyd, L. G. & Vukovic, J. Mate Preferences Across the Lifespan. in The Oxford Handbook of Evolutionary Psychology and Behavioral Endocrinology (eds. Welling, L. L. M. & Schackelford, T. K.) (Oxford University Press, 2018). Little, A. C., Penton-Voak, I. S., Burt, D. M. & Perrett, D. I. Investigating an imprinting-like phenomenon in humans Partners and opposite-sex parents have similar hair and eye colour. Evol. Hum. Behav. 24 , 43–51 (2003). Perrett, D. I. et al. Facial attractiveness judgements reflect learning of parental age characteristics. Proc. R. Soc. London Ser. B-Biological Sci. 269 , 873–880 (2002). Akao, K. A., Adair, L. & Brase, G. L. Parental attachment style, but not environmental quality, is associated with use of opposite-sex parents as a template for relationship partners. Pers. Individ. Dif. (2016) doi:10.1016/j.paid.2016.04.071. Wiszewska, A., Pawlowski, B. & Boothroyd, L. G. Father-daughter relationship as a moderator of sexual imprinting: a facialmetric study. Evol. Hum. Behav. 28 , 248–252 (2007). Bereczkei, T., Gyuris, P. & Weisfeld, G. E. Sexual imprinting in human mate choice. Proc. R. Soc. B 271 , 1129–1134 (2004). Jacob, S., McClintock, M. K., Zelano, B. & Ober, C. Paternally inherited HLA alleles are associated with women’s choice of male odor. Nat. Genet. 30 , 175–179 (2002). Porter, R. H. & Etscorn, F. Olfactory imprinting resulting from brief exposure in Acomys cahirinus. Nature 250 , 732–733 (1974). Nyby, J., Whitney, G., Schmitz, S. & Dizinno, G. Postpubertal experience establishes signal value of mammalian sex odor. Behav. Biol. 22 , 545–552 (1978). Boothroyd, L. G. & Vukovic, J. Mate Preferences Across the Lifespan. in The Oxford Handbook of Evolutionary Psychology and Behavioral Endocrinology (eds. Welling, L. W. & Shackelford, T. K.) (OUP, 2018). Enquist, M., Aronsson, H., Ghirlanda, S., Jansson, L. & Jannini, E. A. Exposure to Mother’s Pregnancy and Lactation in Infancy is Associated with Sexual Attraction to Pregnancy and Lactation in Adulthood. J. Sex. Med. 8 , 140–147 (2011). Saxton, T. K. Experiences during specific developmental stages influence face preferences. Evol. Hum. Behav. 37 , 21–28 (2016). Štěrbová, Z., Tureček, P., Havlíček, J. & Varella Valentova, J. Women learn paternal characteristics used in mate choice during early childhood and adolescence . doi:10.31234/osf.io/fhrp8. Bulygina, E., Mitteroecker, P. & Aiello, L. Ontogeny of facial dimorphism and patterns of individual development within one human population. Am. J. Phys. Anthropol. 131 , 432–443 (2006). Mileva, M., Young, A. W., Jenkins, R. & Burton, A. M. Facial identity across the lifespan. Cogn. Psychol. 116 , 101260 (2020). Hold, B. & Schleidt, M. The Importance of Human Odour in Non‐verbal Communication. Z. Tierpsychol. 43 , 225–238 (1977). Roberts, S. C. et al. Body Odor Similarity in Noncohabiting Twins. Chem. Senses 30 , 651–656 (2005). Allen, C., Havlíček, J., Williams, K. & Roberts, S. C. Evidence for odour-mediated assortative mating in humans: The impact of hormonal contraception and artificial fragrances. Physiol. Behav. 210 , 112541 (2019). Porter, R. H., Cernoch, J. M. & Balogh, R. D. Odor signatures and kin recognition. Physiol Behav 34 , 445–448 (1985). Štěrbová, Z., Tureček, P. & Kleisner, K. Consistency of mate choice in eye and hair colour: Testing possible mechanisms. Evol. Hum. Behav. 40 , 74–81 (2019). Ah-King, M. & Gowaty, P. A. A conceptual review of mate choice: stochastic demography, within-sex phenotypic plasticity, and individual flexibility. Ecol. Evol. 6 , 4607–4642 (2016). Havlíček, J., Fialová, J. & Roberts, S. C. Individual Variation in Body Odor. in Springer Handbook of OdorHandbook of o (ed. Buettner, A.) 949–961 (Springer, 2017). Methods Participants The sample consisted of 67 triads: heterosexual women (mean age = 24.2 years, SD = 3.3), their fathers (mean age = 51.5 years, SD = 6.3) and romantic partners (mean age = 26.7 years, SD = 4.4). The required sample size was estimated based on a priori power analysis in which, based on previous studies, we assumed an association between partner’s and opposite-sex parent’s characteristics to be equivalent to a positive correlation of value 0.25 (for details, see the study’s preregistration at https://osf.io/adrtc). Only women who met the following criteria were enrolled in the study: i) aged 18-35 years, ii) currently in a committed long-term (for a minimum of 6 months) heterosexual romantic relationship (actual mean length = 59 months, SD = 45.3, range = 8-233); iii) shared a common household with their biological father until at least 8 years of age (actual mean = 20.9, SD = 3.6, range = 8-30 years); and iv) not suffering from conditions affecting their sense of smell which may in turn influence their olfactory preferences 31 . The enrolment criteria for fathers and romantic partners were: i) aged under 65 years (fathers) or 18-40 years (partners) ii) non-smokers; iii) in good health and having no underlying condition which affects body odour (e.g., diabetes) 30 ; iv) not currently using medication; and v) having unshaven axillae 32 . Participants were recruited via an e-mail contact list compiled from our previous studies, social media (Facebook, Instagram), and flyers advertised at Charles University, Prague. To check whether participants met the enrolment criteria (e.g., age, health conditions), they completed an online contact form via the Qualtrics platform. Subsequently, we arranged a personal meeting with one individual from each triad and handed over the package with material for data collection. During this meeting, participants were informed about the aims of the study and the sampling procedure was fully explained by going over the written instructions which accompanied the package. Procedure Within the space of one week, each woman completed questionnaires to assess the quality of the relationship with her father during childhood (s-EMBU and overall assessment of childhood relationship quality) and her current attachment style (ECR-RS). Fathers and partners were asked to provide body odour samples for perceptual ratings, chemical analysis, and axillary microbiome analysis. Participants also provided facial photographs and voice recordings which are not the subject of this paper and will be reported elsewhere. Data collection was carried all over the Czech Republic (32 villages, towns or cities, Fig. 1a), increasing the variability of socio-demographic characteristics of our sample (Extended data Table 1). Participants confirmed their participation by signing a consent form. Each triad received compensation for their participation in the sum of 1200 CZK (app. 44 EUR). The study was preregistered (https://osf.io/adrtc) and was approved by the Institutional Review Board of Charles University, Faculty of Science (no. 2017/20). Body odour collection Each father and partner obtained a package including printed instructions and materials for sample collection (non-perfumed liquid soap, tissues, six sterile cotton swabs, three glass vials, three Eppendorf tubes, a vial with the sterile physiological solution, three cotton pads separately enclosed in the labelled zip-lock plastic bags, surgical tape, white cotton T-shirt and a polystyrene transport box (15x20x10 cm). Participants were asked to refrain from consuming aromatic food (a list was provided), drinking alcohol, smoking or using drugs, using cosmetics (e.g., perfumes and deodorants) and from strenuous physical activities including sex during the 48 hours preceding sample collection. All these activities may influence body odour quality 33 . After collection of the samples, each participant completed a survey to document rule compliance (Supplementary Table 1). Although there were some minor violations two days before sampling, there were none reported in the 24h period leading up to the sampling. In the evening on the sampling day, participants collected axillary samples for chemical analysis and then for microbiome analysis. They first washed their hands using non-perfumed liquid soap and dried them with the tissues. To collect samples for chemical analysis, participants wiped each axilla 20 times with a clean cotton swab and placed each swab in a 2 ml empty glass vial. The same procedure was used for collecting axillary microbiome samples, except that the swabs were first soaked in sterile physiological solution. In both types of sampling, they used a sterile cotton swab, unique for each axilla. Axillary microbiome swabs were placed inside an Eppendorf vial containing absolute (99.8%) ethanol. Finally, two control samples (one each for chemical and axillary microbiome analysis) were collected by waving a clean cotton swab 20 times in the air); these samples were used to assess possible effect of environment (e.g., specific odour of the household). All these samples were put into the pre-labelled zip-lock plastic bags and odourless polystyrene box and placed in a freezer (-20°C) immediately after sampling (Fig. 1b). Subsequently, the participants took a shower and washed with the provided non-perfumed liquid soap. They then attached a cotton pad to each axilla using surgical tape. To minimise odour contamination from clothing and other extrinsic ambient odours, participants wore a new white 100% cotton T-shirt as the first layer of their clothing. They wore the pads for the following 12 hours (overnight) 34,35 . Participants collected a further control sample by placing a clean cotton pad on their bedside table for the same period that they wore the axillary pads. Next morning, they placed the cotton pads in separate, pre-labelled zip-lock plastic bags and added these to the polystyrene box together with the samples for chemical and axillary microbiome analysis, which was then put back into the freezer 36 . They returned all samples to the experimenters within a week (in most cases, the experimenters collected these from the participant’s homes). For an overview of the entire procedure, see Supplementary Fig. 1. Questionnaires The quality of relationship with the father during childhood was assessed using the s-EMBU scale (an acronym for ‘Egna Minnen Beträffande Uppofostran’, Swedish for ‘My memories of upbringing’) 37 . This scale is a widely used tool for assessing adults’ perception of their parents’ rearing behaviour. It consists of 23 items, loading onto three sub-scales: Rejection (7 items; possible range 7–28), Emotional Warmth (7 items; possible range 7–28), and (Over)Protection (9 items; possible range 9–36). Responses were indicated on a 4-point Likert-type scale (ranging from 1 – ‘No, never’ to 4 – ‘Yes, most of the time’). We used a Czech language version used in previous studies 38 . We further used a single item to explore overall quality of the relationship during childhood (‘How would you evaluate your relationship with your father during childhood?’), indicated on a 7-point Likert-type scale (ranging from 1 – ‘Very negative’ to 7 – ‘Very positive’). Adult attachment style between the woman and her father was assessed using the Experiences in Close Relationships – Relationship Structures (ECR-RS) questionnaire 39 . This consists of 9 items about attachment orientation of the daughter to father and comprises two subscales: Anxiety (3 items; possible range 3–21) and Avoidance (6 items; possible range 3–42). Daughters completed a Czech version of the scale using a 7-point Likert-type scale (ranging from 1 – ‘Strongly disagree’ to 7 – ‘Strongly agree’). Mean scores for each questionnaire are reported in Extended data Table 1. Perceptual Ratings Raters In total, 306 female raters (mean age = 24.0, SD = 4.0) took part in the rating sessions. They were recruited via an e-mail contact list compiled from our previous studies, social media (Facebook, Instagram) and flyers or by oral invitation around Charles University and the Czech Technical University. Only women in good respiratory health and without any olfactory malfunction took part in the rating sessions. Each rater could participate only in a single rating session. At the end of the session, raters were informed about the study goals and obtained 150 CZK (approximately 5.5 EUR) as compensation for their time. Match-to-sample tests To assess perceived similarity between father and partner body odour we used a match-to-sample paradigm. In total, 67 body odour sets were created. Each set consisted of a father’s body odour sample (the ‘target’), the partner’s body odour sample (the ‘match’) and body odour samples from three other individuals (‘distractors’). The distractor body odours were collected from independent male donors (N= 89, mean age = 25.8; SD = 5.4) who met the same conditions as partners (e.g., the same age, being non-smokers, having unshaven axillae) and who followed the same body odour sampling procedure as described above. Each set consisted only of samples collected from one axillary side (i.e., left/right). In total, 28 sets consisted of left axillary samples and 39 sets used right axillary samples. We sometimes also used partners’ body odour samples as distractors, but these were always the other axillary sample (in other words, if a sample from the left axilla was used as a match, the right axilla sample was used as a distractor). In each session (13 sessions in total), 5 sets of stimuli were used (except the last two sessions consisting of 6 sets). The initial position of the match and the distractors was randomized for each rater (Fig. 1c). The match sample position was then rotated clockwise for each subsequent rater. The distractors used for each rater were randomly chosen from the pool of 15 (or 18 in the last two sessions) distractors available for the given rating session. Similarly, the order of the individual sets was randomized for each rater. Samples were randomized by a blinded experimenter who followed a randomization matrix generated before the rating session. Rating session Rating sessions took place in a quiet, ventilated room with a mean temperature of 22.5°C. The body odour samples were thawed one hour before the rating session 36 . Subsequently, the samples were enclosed in 250ml opaque labelled jars. Raters provided informed consent and were informed about the rating task. The origin of the body odour samples was not disclosed. First, raters were asked to evaluate similarity of the body odours within each set (‘Please sniff the target and then the four other samples. Order the four samples from the most similar to the least similar as compared to the target’). Raters marked their scores into tablet using a purpose-built application MatchApp. The sniffing time was not restricted, most raters finished the odour rating task in 30 minutes. The body odour ratings were followed by assessments of facial photographs and voice recordings, results of which are not relevant to this study. Statistical analysis of perceptual ratings The perceived father-partner similarity in body odour was first evaluated using Chi-square tests - the perceived similarity rankings for each father-partner dyad were compared to the null hypothesis of uniform random distribution of the ranks (i.e., that the match (partner) should appear as the most similar of the four samples in 25 % of cases). Next, to analyse the relative frequency of the match appearing in each of the four ranks, while accounting for non-independence of individual rankings and target’s identity we employed an ordinal probit bayesian mixed model (fitted using R package brms) 40 where parameter estimates and 95% posterior credible intervals were reported. The identity of the raters, the identity of the targets, and axillary side (left/right) were all included as random effects. For a subset of the partner samples, there was no difference between left and right axilla in similarity ratings (estimate [± 95% posterior credibility intervals] = 0.04 ± [-0.43~ 0.50]). The potential effect of relationship quality with the father on perceived father-partner odour similarity was analysed by non-parametric paired Spearman’s correlation (Extended data Table 2). Correlations between individual measures of the relationship quality were generally moderate (Spearman's rho < 0.4) except for the strong negative association between the variables Overall relationship quality and EMBU Rejection (Spearman's rho = -0.59) and the positive association of Overall relationship quality and EMBU Emotional Warmth (Spearman's rho = 0.57) (Supplementary Table 3). Axillary microbiome analysis Metagenomic DNA was extracted using PowerSoil (Qiagen) kits and sequencing libraries were prepared using a two-step PCR approach, following Glenn et al. 41 . Gene-specific primers covering the V3-V4 variable region of bacterial 16S rRNA (i.e., S-D-Bact-0341-b-S-17 [CCTACGGGNGGCWGCAG] and S-D-Bact-0785-a-A-21 [GACTACHVGGGTATCTAATCC]) were used during the first PCR step 42 . The primers were extended at the 5’ end with “tails” serving as a priming site for the second PCR (i.e., the first 33 bp and 34 bp of read1 and read2 Nextera adapters for F and R primers, respectively). The first PCR was performed in 10 μl and consisted of 5 μl 1x KAPA HiFi Hot Start Ready Mix (Roche), each primer at 0.2 μM and 4.6 μl DNA. PCR conditions were as follows: initial denaturation at 95°C for 3 min followed by 35 cycles each of 95°C (30 s), 55°C (30 s), and 72°C (30 s), and a final extension at 72°C (5 min). Dual-indexed Nextera sequencing adapters were reconstructed during the second PCR, which followed the first PCR conditions except that: it was performed in 20 μl volume; the concentration of each primer was 1 uM; 6μl of the first PCR product were used as a template; and the number of PCR cycles was 12. Technical duplicates were prepared to account for noise due to PCR and sequencing stochasticity. Products of the second PCR were run on 1.5% agarose gel and their concentrations were assessed based on gel band intensities using GenoSoft software (VWR International, Belgium). They were pooled equimolarly and size-selected with Pippin Prep (Sage Science, USA) at 520 - 750 bp. Resulting libraries were sequenced using MiSeq (Illumina, USA) and v3 chemistry (i.e., 2 × 300 bp paired-end reads). Sequencing data are available at the European Nucleotide Archive under the accession number of the whole project ( will be completed upon the acceptance of the manuscript ). Samples were demultiplexed and primers were trimmed by skewer software 43 . Using dada2 44 , we eliminated low-quality sequences (expected number of errors > 2), denoised quality-filtered fastq files, and constructed abundance matrix representing read counts for individual amplicon sequencing variants (hereafter ASV) in each sample. Next, we identified chimeric ASVs using uchime 45 and gold.fna reference database and eliminated them from the abundance matrix. Taxonomic assignation of ASVs was conducted by RDP classifier (80% confidence threshold 46 ) and Silva database (v. 132) 47 . Technical replicates of axillary microbiota samples exhibited high consistency in both composition (Procrustean correlation, r = 0.99, p = 0.0001) and Shannon diversity (Pearson's r = 0.93, p < 0.0001). Consequently, we merged sequences corresponding to individual samples for all subsequent analysis. To avoid microbial diversity inflation due to PCR and sequencing artifacts, we eliminated all OTUs that were present just once within the set of PCR replicates for a given sample 48 . Prior to statistical calculations, we also excluded one sample that exhibited low consistency among its technical replicates and one sample of low sequencing coverage (< 1500 high quality sequences). The final dataset included 246 samples of left and/or right axillary microbiota coming from 132 individuals (66 fathers and 66 partners). These samples were represented by 3,129,530 high-quality reads that were distributed among 4,891 ASVs. The median sequencing coverage was 13,152 reads per sample (range = 1,750 – 34,343). Functional predictions of bacterial metagenome were conducted by PICRUSt2 pipeline 49 using the default setup. Predicted metagenomes were then categorized to functional pathway 50 and their predicted abundances in each sample were used in later statistical analyses. Weighted Nearest Sequenced Taxon Index (weighted NSTI) was 0.02755007 +/- 0.001274799 (+/- S.E) suggesting very high precision of metagenome predictions. In addition to the axillary microbiota samples, we also collected 74 environmental microbiota samples. Due to lower DNA concentrations and PCR stochasticity, duplicates from the environmental samples had lower alpha diversity (r = 0.74, p < 0.0001) and compositional consistency (r = 0.68, p 1000 reads after all filtering steps. Microbial alpha diversity did not significantly vary between left and right armpit (Generalized Linear Mixed Model (GLMM): Δ D.f. = 1, χ2= 1.31, p = 0.252 for ASVs richness and Δ D.f. = 1, χ2= 1.31, p = 0.252 for Shannon diversity) and there was also no significant difference in the number of detected ASVs between fathers and partners (GLMM: Δ D.f. = 1 χ2= 2.06, p = 0.151). However, partners exhibited slightly lower Shannon diversity compared to fathers (GLMM: estimate = -0.24 +/- 0.12, Δ D.f. = 1, χ2= 4.27, p = 0.039). NMDS (Extended Data Fig. 1a) and PERMANOVA analyses did not detect any systematic difference in the composition between left and right axilla, yet they revealed a slight compositional shift between fathers and partners explaining only < 2% of total variation in composition (Extended Data Table 3a and Extended Data Fig. 1b). Similarly, alpha diversity was closely correlated in the left and right axilla (Extended Data Fig 1c). According to GLMM analyses with distances to group-specific centroids as a response, profiles of left and right axilla were equally dispersed, but fathers exhibited higher dispersion in composition (i.e., inter-individual variation) than partners (Extended Data Table 3b and Extended Data Fig. 1d). Samples from both left and right axillae were successfully profiled in the case of 114 individuals (58 fathers and 56 partners). Based on this subset, there was a tight correlation in alpha diversity between left and right axillae (Pearson’s correlation: r = 0.796, p < 0.0001 for ASVs richness and 0.797 for Shannon diversity). The same holds for consistency in the composition, where microbiota divergences calculated using ASVs presence/absence was more decisive (Procrustean correlation for Jaccard distances: r = 0.94, p = 0.0001) than divergence metrics utilizing relative ASVs abundances (Procrustean correlation for Bray-Curtis distances: r = 0.82, p = 0.00001, Fig. 2b and Extended Data Fig. 1c). Furthermore, consistency between left and right axillae was significantly higher for fathers than for partners (GLMM with Procrustean residuals as a response: Δ D.f. = 1, χ2= 5.44, p = 0.02 for Bray-Curtis and Δ D.f. = 1, χ2= 9.09, p = 0.003 for Jaccard dissimilarities, Extended Data Fig. 1d). For the subsequent analyses, profiles of left and right axillae of the same individual were aggregated. Statistical analysis of microbiome data Shannon indices and number of ASVs in each sample (hereafter ASVs richness) were calculated after dataset normalization by rarefaction (threshold = 1,750 reads per sample) and used as alpha diversity measures. We analysed differences in alpha diversity between fathers and partners and left vs. right axillae using GLMM with Gaussian error distribution. Individual identity nested within pair identity were specified as random effects. Pearson correlation was used to check for a correlation in alpha diversity between left and right axillae. Differences in composition of microbiota profiles was analysed based on Bray Curtis dissimilarities accounting for ASV relative abundances and binary Jaccard dissimilarities accounting just for ASVs presence/absence. Both these metrices were calculated after the rarefaction of the ASV abundance matrix. First, we checked if there were any differences in composition between left and right axillae and between fathers and partners using Nonmetric Multidimensional Scaling (NMDS) and PERMANOVA with permutation constraint (i.e., ‘strata’) setup to within-pair level. To test if interindividual variation differed between left and right axillae, and between fathers and partners, we exacted distance to centroids for father’s and partner’s samples and used them after log transformation as a response in GLMM, assuming Gaussian errors and the same random structure as described above. ASVs overrepresented in fathers or partners were identified by differential abundance analyses, where read counts for each ASV were included as a response into GLMM with negative binomial errors and log-scaled total sequencing depth for a given sample as an offset. The model random structure remained unchanged. To achieve model convergence, these models were fitted using glmmTMB package 51 only for ASVs detected in at least 10 samples. The Qvalue method 52 was then used to control for false discoveries due to multiple testing, with the threshold being setup to q < 0.05. The composition and alpha diversity of left and right axillary microbiota were tightly correlated at the individual level. We therefore aggregated the left and right axillary profiles for each individual and used the aggregated profiles to test if there was higher similarity between father-partner pairs than expected by chance. To do this, we compared average microbiota dissimilarity between actual father-partner dyads with a distribution of average dissimilarities drawn from randomly paired fathers and partners (n = 999 permutations). The outcome of these analyses were also expressed as Standardised Effect Sizes (SES) 53 using the formula SES = (∑DIF (obs) – ∑DIF (sim) /nperm)/sd(∑DIF (sim) ), where ∑DIF (obs) is the sum of observed within dyad dissimilarities, ∑DIF (sim) is the sum of simulated within-dyad dissimilarities, nperm is number of permutations, and sd is standard deviation. A simulation study was conducted to test the robustness of this permutation routine to false positives. First, using the R package AHMbook 54 , we simulated 1,000 random communities of 1,000 species mimicking father-partner dyads and thus consisting of two sample groups of 60 samples each. No a prior i driver of correlation in community composition was specified within the 60 dyads. The frequency of false positive results estimated, based on these simulated datasets and the permutation routine described above, was low at 4.6% (Extended Data Fig. 1e). Analyses on predicted metagenome pathways utilized most of the above-described approaches, with two distinctions. First, instead of abundance matrix rarefaction, functional pathways abundances were normalized by their transformation to proportions in each sample. Second, we used only relative-based (i.e., Bray-Curtis) dissimilarities in all analyses on predicted metagenomes. All statistical calculations and data visualizations were conducted using R 4.0.2 55 . Chemical analysis Sample extraction The sampling for analyses of axillary chemical profiles was performed by the participants themselves as described above. The samples were transported and stored at -20°C until extraction, which took place as soon as possible after the reception of samples. Left and right axillary samples were successfully collected from all participants (67 father-partner dyads). To allow multiple analyses and storage of the irreplaceable samples, we selected the solvent extraction of the samples, in spite of the drawbacks inherent to this method when compared with direct thermal desorption, especially the lower representation of highly volatile and low molecular mass compounds. The analytes were extracted from the cotton swabs with 400 µl of distilled n -hexane for 24 hours at 8°C plus an additional 10 minutes of sonication. The extracts were transferred into clean vials, two internal standards (10 ng of 1-bromononane and 100 ng of 1-bromoeicosane) were added, and the extracts were either stored at -80°C or directly analysed. Instrument setup The samples were analysed using a GC×GC-TOFMS instrument Pegasus 4D (LECO Corp., St Joseph, MI, USA). The system is based on a quad-jet modulator with two cold liquid-nitrogen jets and two hot-air jets, which trap and refocus the analytes eluting from the first dimension column. The column set consisted of a nonpolar Rxi-5Sil MS (30 m, ID 0.25 mm, d f 0.25 µm, Restek, Bellefonte, PA, USA) in the first dimension and a mid-polar Rxi-17Sil MS (1.5 ms, id 0.1 mm, d f 0.1 µm, Restek, Bellefonte, PA, USA) in the second dimension. The columns were connected using SilTite® µ-Union (Trajan Scientific, Australia). The modulation period was 5 s and the hot pulse duration 600 ms. The temperature program was 50°C (1 min), 8°C/min ramp to 320°C (20 min). The secondary column was set 10°C higher. The modulator offset was 20°C. Helium was used as a carrier gas at a constant flow rate of 1 ml/min. Total run time was 54.75 min. 1µl of the extract was injected into a split/splitless injector heated to 200°C in splitless mode. Acquisition solvent delay was set at 400 s, while MS transfer line temperature was 280°C. The ion source temperature was 220°C and electron impact energy was -70 eV. Collected mass range was 29-600 amu and acquisition rate was 100 scans/s. The detector voltage was 1,500 V. Measurement sequence The samples were analysed in randomized order and injected using a 7683 Series Injector (Agilent Technologies). Before each sample, a blank was run using the same temperature program as for the sample runs. To guarantee the reliability of the chemical analyses and to avoid biases in chromatograms alignment, we only considered the samples which were analysed during a single measurement sequence, uninterrupted by any change in the analytical setup, such as column exchange, etc. The only intervention in the setup during the sequence measurement was the regular exchange of the split/splitless liner (Topaz liner, Splitless, Single Taper Gooseneck w/Wool, 4mm × 6.5 ×78.5, Restek), which was replaced after every 10 samples. The longest successful uninterrupted sequence included both left and right axillary samples of 42 father-partner dyads (168 samples), which were then used for data analysis. Data processing For peak detection, alignment and quantification, we processed the data generated by the operating software LECO ChromaTOF software v. 4.72 (LECO, St. Joseph, MI) with the development version of ChromaTOF-Tile (v. 0.27). This software applies the recently developed tile-based algorithm for GC×GC-TOFMS multiple chromatogram processing, which provides an effective alternative to the widespread pixel-based and peak table-based methods 56,57 . Tile-based algorithm offers two main advantages over the traditional approaches: it is computationally fast and it significantly reduces the problem of 2D retention time misalignments. ChromaTOF-Tile is primarily designed for supervised comparison of two or multiple pre-defined classes (groups) of samples. Therefore, for the purposes of unsupervised exploration of chemical profiles across all body odour samples with no à priori classification, we defined all measured samples as one group of 168 samples. To generate the second, reference (control) group, we used the chromatogram of a hexane solution of n-alkanes (C 7 ‒C 40 , 1ng/µl each, Sigma Aldrich), obtained using identical chromatographic conditions as the samples. Upon importing into ChromaTOF-Tile, we multiplicated this chromatogram into 16 files with variable dilution factors (0.96‒1.03) to give rise to the reference group with identical and low coefficients of variation for all analytes. To reduce the number of multiple thousand putative analytes initially detected by ChromaTOF-Tile, operated over the range of masses m/z 29‒600, to the most informative peaks characterizing body odour samples, we took advantage of F-ratio comparison between the two predefined classes calculated by ChromaTOF-Tile for each analyte’s peak volumes. At the same time, we optimized the signal-to-noise ratio (S/N ratio) for peak inclusion to eliminate artifact peaks. Both parameters (i.e., F-ratio and S/N ratio) were fine-tuned during the pilot study with six male participants sampled six times each (Supplementary information). Desirable separation of chemical profiles of the six participants and optimum clustering of the six samples per participant were obtained for F-ratio = 2 and S/N ratio = 50 (Extended data Fig. 5). These inclusion parameters were thus used for the main analysis. The resulting dataset was then manually curated to exclude the remaining column bleeding signals and artifacts, peak tailing-related hits, and analytes identified in the extracts of the holding sticks of the cotton swabs used for sampling. Analytes identification For tentative identification of individual analytes included in the main analysis, both MS spectra and 2D retention characteristics were used. Electron ionization MS spectrum (EIMS) of each compound was compared with mass spectral libraries NIST 2017, Wiley Registry™ of Mass Spectral Data (8 th edition) and an in-house wax ester library 58 . EIMS spectra of all analytes were manually inspected, eventual erroneous library hits corrected, and new identifications proposed based on diagnostic fragments. In parallel, retention indices (RI) for both GC dimensions were retrieved from ChromaTOF-Tile output data and manually compared with NIST and other sources 59,60 . Tentative identification is provided only for compounds matching by both the EIMS and RI data (Supplementary Table Dataset 2). Statistical analysis The peak volumes for all included analytes across the 168 samples were square root transformed to reduce heteroscedasticity and related to the sum of peak volumes of each sample to correct for the differences in sample intensities. For subsequent calculations, three datasets with 84 samples each were prepared and independently processed. The first two included only right or only left axillary samples, while the third consisted of averaged values from left and right axillary sample for each analyte. Each dataset was subjected to Principal Component Analysis with autoscaling and centering, performed in Canoco for Windows v. 4.52. The factor scores for the first three principal components were plotted and used to calculate the matrix of Euclidean distances among all possible pairs of participants. These distances were then used as indicators of chemical dissimilarity among different samples. The differences between within-dyad father-partner distances and averaged distances of the corresponding father to all extra-pair partners were compared with zero expected under the null hypothesis of no influence of the pairing on chemical similarities, using Wilcoxon Signed Rank tests. Independently, we tested whether the observed within-dyad father-partner distances differ from distances obtained under random pairing of fathers with partners. To do so, we generated datasets of randomly paired father and partners (999 permutations) and compared the observed average father-partner distance with the distribution of average father-partner distances in the permuted datasets. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. 31. Schäfer, L., Schriever, V. A. & Croy, I. Human olfactory dysfunction: causes and consequences. Cell Tissue Res. 383 , 569–579 (2021). 32. Kohoutová, D., Rubešová, A. & Havlíček, J. Shaving of axillary hair has only a transient effect on perceived body odor pleasantness. Behav. Ecol. Sociobiol. 66 , 569–581 (2012). 33. Havlicek, J. & Lenochova, P. Environmental Effects on Human Body Odour. in Chemical Signals in Vertebrates XI (eds. Hurst, J. L., Beynon, R. J., Roberts, S. C. & D., W. T.) vol. 11 199–212 (Springer, 2008). 34. Fialová, J., Roberts, S. C. & Havlíček, J. Consumption of garlic positively affects hedonic perception of axillary body odour. Appetite 97 , 8–15 (2016). 35. Havlíček, J., Lenochová, P., Oberzaucher E., Grammer, K. & Roberts, S. C. Does length of sampling affects quality of body odor samples? Chemosens. Percept. (2011). 36. Lenochova, P., Roberts, S. C. & Havlicek, J. Methods of Human Body Odor Sampling: The Effect of Freezing. Chem. Scences 34 , 127–138 (2009). 37. Arrindell, W. A. et al. The development of a short form of the EMBU 1: Its appraisal with students in Greece, Guatemala, Hungary and Italy. Pers. Individ. Dif. 27 , 613–628 (1999). 38. Šaffa, G., Štěrbová, Z. & Prokop, P. Parental Investment Is Biased toward Children Named for Their Fathers. Hum. Nat. 32 , 387–405 (2021). 39. Fraley, R. C., Heffernan, M. E., Vicary, A. M. & Brumbaugh, C. C. The Experiences in Close Relationships-Relationship Structures Questionnaire: A Method for Assessing Attachment Orientations Across Relationships. Psychol. Assess. 23 , 615–625 (2011). 40. Bürkner, P. C. brms: An R package for Bayesian multilevel models using Stan. J. Stat. Softw. 80 , 1–28 (2017). 41. Glenn, T. C. et al. Adapterama I: universal stubs and primers for 384 unique dual-indexed or 147,456 combinatorially-indexed Illumina libraries (iTru & iNext). PeerJ 7 , e7755 (2019). 42. Klindworth, A. et al. Evaluation of general 16S ribosomal RNA gene PCR primers for classical and next-generation sequencing-based diversity studies. Nucleic Acids Res. 41 , 1–11 (2013). 43. Jiang, H., Lei, R., Ding, S. W. & Zhu, S. Skewer: A fast and accurate adapter trimmer for next-generation sequencing paired-end reads. BMC Bioinformatics 15 , 1–12 (2014). 44. Callahan, B. J. et al. DADA2: High-resolution sample inference from Illumina amplicon data. Nat. Methods 13 , 581–583 (2016). 45. Edgar, R. C., Haas, B. J., Clemente, J. C., Quince, C. & Knight, R. UCHIME improves sensitivity and speed of chimera detection. Bioinformatics 27 , 2194–2200 (2011). 46. Wang, Q., Garrity, G. M., Tiedje, J. M. & Cole, J. R. Naïve Bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy. Appl. Environ. Microbiol. 73 , 5261–5267 (2007). 47. Quast, C. et al. The SILVA ribosomal RNA gene database project: Improved data processing and web-based tools. Nucleic Acids Res. 41 , 590–596 (2013). 48. Pafčo, B. et al. Metabarcoding analysis of strongylid nematode diversity in two sympatric primate species. Sci. Rep. 8 , 1–11 (2018). 49. Douglas, Gavin, M. et al. PICRUSt2: An improved and extensible approach for metagenome inference. bioRxiv 672295 (2020). 50. Ye, Y. & Doak, T. G. A parsimony approach to biological pathway reconstruction/inference for genomes and metagenomes. PLoS Comput. Biol. 5 , 1–8 (2009). 51. Brooks, M. E. et al. glmmTMB balances speed and flexibility among packages for zero-inflated generalized linear mixed modeling. R J. 9 , 378–400 (2017). 52. Storey, J. D. & Tibshirani, R. Statistical significance for genomewide studies. Proc. Natl. Acad. Sci. U. S. A. 100 , 9440–9445 (2003). 53. Gotelli, N. J. & McCabe, D. J. Species co-occurrence: A meta-analysis of J. M. Diamond’s assembly rules model. Ecology 83 , 2091–2096 (2002). 54. Kery, M. & Royle, J. A. Applied Hierarchical Modeling in Ecology: Analysis of distribution, abundance and species richness in R and BUGS: Volume 2: Dynamic and Advanced Models. (Academic Press, 2020). 55. R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing. (2020). 56. Marney, L. C. et al. Tile-based Fisher-ratio software for improved feature selection analysis of comprehensive two-dimensional gas chromatography-time-of-flight mass spectrometry data. Talanta 115 , 887–895 (2013). 57. Parsons, B. A. et al. Tile-Based Fisher Ratio Analysis of Comprehensive Two-Dimensional Gas Chromatography Time-of-Flight Mass Spectrometry (GC × GC-TOFMS) Data Using a Null Distribution Approach. Anal. Chem. 87 , 3812–3819 (2015). 58. Urbanová, K., Vrkoslav, V., Valterová, I., Háková, M. & Cvačka, J. Structural characterization of wax esters by electron ionization mass spectrometry. J. Lipid Res. 53 , 204–213 (2012). 59. Adams, R. P. Identification of Essential Oil Components by Gas Chromatography/mass Spectrometry . (Allured Publishing Corporation, 2007). 60. Stránský, K., Zarevúcka, M., Valterová, I. & Wimmer, Z. Gas chromatographic retention data of wax esters. J. Chromatogr. A 1128 , 208–219 (2006). Additional Declarations There is NO Competing Interest. Supplementary Files HavlicekSupplement.pdf Supplementary Information Havlicekperceptualratings.xlsx Dataset 1 Havlicekchemicalanalyses.xlsx Dataset 2 Extendeddata.doc Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-2531352","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Social Sciences - Article","associatedPublications":[],"authors":[{"id":172200912,"identity":"646496e0-ecbc-4b88-ba2f-cbd257de2835","order_by":0,"name":"Jan Havlíček","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA80lEQVRIiWNgGAWjYBACxhlA4oEBA4MEewNC9ABBLQkgLTwHEErxamGQAGkB0RIJhJRCAfPs5mcSCQV2cpIz3x78/LHtHgP/jNyDBxhq6nA7bM4xM4kEg2Rjaem8ZImDbcUMEjfyEg4wHGPD45cEY4MEgwOJ86RzDIBaEhgMJHIMDjA28ODRkv4ZpKV+nuQZ4x9IWiTwaMkxfADUkiAtwWOGbIsBPi2FQC3JhjN7cswszpxL4JE48wZowrEEnFoMZ6RvOPDhj528xPEzxjcqyhLk+NtzjD98wBNihg1oAhBv47aDgUEej9woGAWjYBSMAggAANbXUxoLCA11AAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0003-2374-7337","institution":"Charles University","correspondingAuthor":true,"submittingAuthor":false,"prefix":"","firstName":"Jan","middleName":"","lastName":"Havlíček","suffix":""},{"id":172200913,"identity":"5a539907-f119-4097-850f-fa1cacb5b059","order_by":1,"name":"Lucie Jelínková","email":"","orcid":"","institution":"Department of Zoology, Faculty of Science, Charles University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Lucie","middleName":"","lastName":"Jelínková","suffix":""},{"id":172200914,"identity":"2651ef01-f8e2-4b7e-b487-4629bdbf8df8","order_by":2,"name":"Zuzana Štěrbová","email":"","orcid":"","institution":"Department of Zoology, Faculty of Science, Charles University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Zuzana","middleName":"","lastName":"Štěrbová","suffix":""},{"id":172200915,"identity":"88447d7e-2569-485d-a571-f737e0eacfe5","order_by":3,"name":"Robert Hanus","email":"","orcid":"","institution":"Institute of Organic Chemistry and Biochemistry Academy of Sciences of the Czech Republic","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Robert","middleName":"","lastName":"Hanus","suffix":""},{"id":172200916,"identity":"7a055b23-fd7e-4995-8dd2-1bbd6b6353d8","order_by":4,"name":"Jakub Kreisinger","email":"","orcid":"","institution":"Department of Zoology, Faculty of Science, Charles University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Jakub","middleName":"","lastName":"Kreisinger","suffix":""},{"id":172200917,"identity":"a5025cad-730e-4f39-a3e8-be6621aa3132","order_by":5,"name":"Pavlina Kyjakova","email":"","orcid":"","institution":"Institute of Organic Chemistry and Biochemistry, Academy of Sciences of the Czech Republic","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Pavlina","middleName":"","lastName":"Kyjakova","suffix":""},{"id":172200918,"identity":"20c82e6c-821a-45f9-8583-1acb6a9673da","order_by":6,"name":"Lucie Schmiedová","email":"","orcid":"","institution":"Department of Zoology, Faculty of Science, Charles University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Lucie","middleName":"","lastName":"Schmiedová","suffix":""},{"id":172200919,"identity":"930b167c-5ba8-4cde-a9e2-5e30d424f762","order_by":7,"name":"Radka Bušovská","email":"","orcid":"","institution":"Institute of Organic Chemistry and Biochemistry Academy of Sciences of the Czech Republic","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Radka","middleName":"","lastName":"Bušovská","suffix":""},{"id":172200920,"identity":"8dae3e9a-4830-4f7b-ac68-f17b17ea58e1","order_by":8,"name":"Jitka Třebická Fialová","email":"","orcid":"","institution":"Department of Zoology, Faculty of Science, Charles University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Jitka","middleName":"Třebická","lastName":"Fialová","suffix":""},{"id":172200921,"identity":"5fa60de3-bb86-4eda-8d6d-d439d646cf26","order_by":9,"name":"Dagmar Schwambergová","email":"","orcid":"","institution":"Department of Zoology, Faculty of Science, Charles University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Dagmar","middleName":"","lastName":"Schwambergová","suffix":""},{"id":172200922,"identity":"debb393d-fcc9-45b9-89ac-c52c25b80e27","order_by":10,"name":"Craig Roberts","email":"","orcid":"","institution":"","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Craig","middleName":"","lastName":"Roberts","suffix":""}],"badges":[],"createdAt":"2023-01-30 21:41:32","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-2531352/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-2531352/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":32289409,"identity":"a0f16c24-8f62-4cda-a5a6-899f8d748aae","added_by":"auto","created_at":"2023-01-31 19:34:30","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":333547,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSampling procedure and perceived similarity.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea, \u003c/strong\u003eMap of the Czech Republic showing the breadth of sampling locations; the size of the red circles represents the number of participants collected in a given location. \u003cstrong\u003eb,\u003c/strong\u003e Procedure of sample collection for axillary microbiome and chemical analysis (swabs taken, upper pathway) and perceptual task (axillary pads, lower pathway). It starts with the experimenter explaining the sampling procedure and outlining lifestyle restrictions to be followed; participants subsequently take axillary swabs and wear pads overnight, each of these samples being stored in the freezer until they are collected by the experimenter. \u003cstrong\u003ec,\u003c/strong\u003e An illustrative stimulus set of body odour samples for the match-to-sample test; target = father sample, match = partner sample and three distractor samples; during presentation to panellists, these labels were coded using geometric images for match and distractor samples and with the number for target sample. \u003cstrong\u003ed, \u003c/strong\u003eResults of perceived similarity rating. Bars show the frequency with which partner body odour samples were ranked from most (1) to least (4) similar to the target father’s odour. Partner samples were evaluated as most similar to the father’s sample significantly more often (χ2 = 92.84; p \u0026lt; 0.001) than expected by chance (25%, marked by dashed line). \u003cstrong\u003ee, \u003c/strong\u003eScatterplots showing\u003cstrong\u003e \u003c/strong\u003ethe relationship between the average perceived partner similarity and measures of the daughters’ relationship quality with their father in childhood: s-EMBU scales (Emotional Warmth, Overprotection, Rejection), overall relationship quality with father during childhood, and in adulthood: ECR-RS scales (Avoidance, Anxiety).\u003c/p\u003e","description":"","filename":"Fig1.png","url":"https://assets-eu.researchsquare.com/files/rs-2531352/v1/61f1db35dca987eb6f9eccc0.png"},{"id":32289769,"identity":"ad60ca7d-e7d2-4293-b6e8-4d3f8871bd5e","added_by":"auto","created_at":"2023-01-31 19:42:30","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":357430,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSimilarity between fathers and partners in the axillary microbiota.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea,\u003c/strong\u003e Relative abundances of the dominant bacterial genera in the axillary microbiota. Father-partner dyad identity and left (l) or right (r) axillary samples are indicated by labelling of horizontal facets. Blank spaces indicate samples that were not included in the final analyses due to low data quality. \u003cstrong\u003eb,\u003c/strong\u003e Similarity in left and right axillary microbiota composition for individual participants visualised by Procrustes superimposition. \u003cstrong\u003ec,\u003c/strong\u003e Non-metric multidimensional scaling (NMDS), with lines highlighting similarities in microbiota composition between father and partner within each dyad. \u003cstrong\u003ed,\u003c/strong\u003e According to a permutation test, the average value of father-partner microbiota dissimilarity in microbiota composition within actual dyads (arrow) was significantly lower than the null distribution of average dissimilarities of randomly paired fathers and partners (histogram). Figures b-d show the patterns of microbiota variation assessed based on Jaccard dissimilarities (accounting for the presence/absence of bacterial ASVs). Results for Bray-Curtis dissimilarities (accounting for the relative abundances of bacterial ASVs) are provided in the Extended Data Fig 2e.\u003c/p\u003e","description":"","filename":"Fig2.png","url":"https://assets-eu.researchsquare.com/files/rs-2531352/v1/2b1bb11de4abcb7aa49bf0b6.png"},{"id":32289410,"identity":"ec2c7572-1077-4a39-aa65-7a472e951862","added_by":"auto","created_at":"2023-01-31 19:34:30","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":481295,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eChemical similarity of axillary odour samples.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea,\u003c/strong\u003e Selected GC×GC chromatograms depicting the chemical profiles of left axillary samples in two father-partner dyads, showing within-dyad similarity even upon visual inspection. \u003cstrong\u003eb, c, \u003c/strong\u003eFactor scores of the first three principal components for 82 fathers and partners, calculated from peak intensities of 341 analytes detected and quantified using tile-based analysis of left-right averaged GC×GC profiles of axillary odour samples. Prior to principal component analysis, the peak volumes were square-root transformed, related to the sum of all 341 peak volumes in each sample, the relative values averaged between left and right axillary samples for each individual, autoscaled and centered. Left graph (b) shows the overall distribution of the 82 observations and the 95% confidence ellipses for fathers (green) and partners (blue). The right graph (c) shows the within-dyad fathers and partners connected with a grey line representing the Euclidean distance. \u003cstrong\u003ed, \u003c/strong\u003eObserved average within-dyad father-partner Euclidean distance (arrow) and null distribution of distances obtained by random pairing of fathers and partners reveal highly significant within-dyad similarity than expected by chance (999 permutations, p = 0.001) for left-right averaged axillary samples. \u003cstrong\u003ee, \u003c/strong\u003eObserved within-dyad father-partner distances subtracted from averaged distances of the corresponding father to all extra-pair partners show a distribution significantly greater than zero for left-right averaged samples (Wilcoxon Signed Rank Test, p = 0.0006).\u003c/p\u003e","description":"","filename":"Fig3.png","url":"https://assets-eu.researchsquare.com/files/rs-2531352/v1/47255b5f3b6b28afba685e72.png"},{"id":32362794,"identity":"eee7cea1-b3a9-42a6-8b13-0aa57887ade9","added_by":"auto","created_at":"2023-02-02 05:30:58","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1685553,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-2531352/v1/f1bccc55-8bae-449f-a5dd-2037b50f92ed.pdf"},{"id":32289413,"identity":"db3c77d6-9e6c-4d63-a355-d53d2966940c","added_by":"auto","created_at":"2023-01-31 19:34:30","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":231049,"visible":true,"origin":"","legend":"Supplementary Information","description":"","filename":"HavlicekSupplement.pdf","url":"https://assets-eu.researchsquare.com/files/rs-2531352/v1/ff219819e2bc34f17a439315.pdf"},{"id":32289411,"identity":"cdb98b95-a3ec-4c11-b01e-10f3dbe7a7a9","added_by":"auto","created_at":"2023-01-31 19:34:30","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":316937,"visible":true,"origin":"","legend":"Dataset 1","description":"","filename":"Havlicekperceptualratings.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-2531352/v1/758f39fbff48f761a108d4d7.xlsx"},{"id":32289415,"identity":"a0702ec7-8ca9-4df0-8054-192598b9d9b9","added_by":"auto","created_at":"2023-01-31 19:34:31","extension":"xlsx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":1049744,"visible":true,"origin":"","legend":"Dataset 2","description":"","filename":"Havlicekchemicalanalyses.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-2531352/v1/05280c55ed684f96e57c546d.xlsx"},{"id":32289414,"identity":"7f60ccb9-2108-4d51-8b0f-97a23d291613","added_by":"auto","created_at":"2023-01-31 19:34:31","extension":"doc","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":3317248,"visible":true,"origin":"","legend":"","description":"","filename":"Extendeddata.doc","url":"https://assets-eu.researchsquare.com/files/rs-2531352/v1/535b99a08ad0e09d410c0b05.doc"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Women choose romantic partners resembling their father in body odour","fulltext":[{"header":"Introduction","content":"\u003cp\u003eCross-fostering studies in mammals show that learnt olfactory cues in the postnatal environment affect subsequent mate choice in adulthood, leading to phenotypic similarity between adoptive parents and the partners of their adoptive offspring\u003csup\u003e2,5\u003c/sup\u003e. What makes this observation even more remarkable, compared to visual or acoustic forms of phenotype-matching, is that such odour-mediated behaviour is not an entirely autonomous process, but appears to be critically dependent on the commensal microflora. Thus, bacterial action on less volatile molecules produced by the mammalian host is required to generate the volatile metabolites which form the host\u0026rsquo;s odour cues\u003csup\u003e6\u0026ndash;8\u003c/sup\u003e. If this is the case, then we would hypothesise that in a species where parental odour characteristics influence mate preferences, we should observe similarities not only in how parental and mate odours are perceived, but also in the microbial communities that are responsible for the odour\u0026rsquo;s generation and the resulting chemical profiles.\u003c/p\u003e\n\u003cp\u003eHere we test this hypothesis for the first time, addressing human mate preferences which appear to form during childhood and where parental characteristics play a pivotal role\u003csup\u003e9\u003c/sup\u003e. For example, women tend to choose partners resembling their father in various physical\u003csup\u003e10,11\u003c/sup\u003e and personality\u003csup\u003e12\u003c/sup\u003e characteristics. Furthermore, levels of facial similarity are positively modulated by the quality of relationship with the father during childhood\u003csup\u003e13,14\u003c/sup\u003e. Some evidence also suggests that women prefer odours of men who have HLA alleles in common with their father\u003csup\u003e15\u003c/sup\u003e, but whether this affects actual partner choice or leads to odour similarity between women\u0026rsquo;s partners and their fathers has not yet been addressed. To answer such questions, we investigated similarity between fathers and the romantic partners of their daughters. For the first time in any species, we did this at each of the three main steps in the hypothesised pathway, examining i) perceptual ratings of odour similarity, ii) similarity in their microbiomes, and iii) similarities in their chemical profiles. \u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePerceived similarity\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo test perceived similarity in body odour we recruited 67 male dyads (a woman\u0026rsquo;s long-term romantic partner and her father) (Fig. 1a). All women grew up in the same household as their father at least up to 8 years of age. The participants refrained from using personal cosmetics, eating aromatic food and from other activities affecting body odour for 2 days preceding the sampling and subsequently collected their axillary body odour samples on cotton pads worn for 12h (overnight) (Fig. 1b). Perceived similarity between the target father sample and the partner was tested by a 4-choice match-to-sample test\u003csup\u003e14\u003c/sup\u003e (Fig. 1c), in which female panellists (unacquainted with the odour donors) were asked to smell and then order the 4 alternatives (the partner and three age-matched distractors) based on their similarity to the target father sample. Each panellist assessed 5\u0026mdash;6 sets of stimuli and each sample was assessed by at least 20 independent panellists (N = 306) which makes 1566 match-to-sample tests in total.\u003c/p\u003e\n\u003cp\u003ePanellists selected the partner\u0026rsquo;s body odour as the most similar to the father\u0026rsquo;s body odour in 35.5 % of cases, which is significantly more frequent than chance (25 %, Chi-square = 92.84, p \u0026lt; 0.001). Furthermore, the frequency of selection in the second-, third- or least-similar position decreased in a near-linear fashion (Fig. 1d). To analyse all these selections while accounting for non-independence of individual ratings and target\u0026rsquo;s identity, we used an ordinal probit mixed model. This analysis revealed substantial differences in similarity ratings between partner and non-partner samples (estimate [\u0026plusmn; 95% posterior credibility intervals] = -0.48 \u0026plusmn; [-0.64 ~ -0.32]). According to the parameter estimates, partner samples had a much greater chance of being rated as most similar or second most similar to the target father sample than the non-partner samples. Conversely, non-partner samples were more likely to be classified as the least similar or the second least similar.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMicrobiome similarity\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe axillary microbiota was analysed by amplicon sequencing of the 16S rRNA gene. The resulting microbiota profiles revealed a dominance of bacteria belonging to the genera \u003cem\u003eStaphylococcus \u003c/em\u003e(42.2% of sequences), \u003cem\u003eCorynebacterium\u003c/em\u003e (31.1% of sequences) and \u003cem\u003eAnaerococcus \u003c/em\u003e(10.0% of sequences, Fig. 2a). Microbial alpha diversity, composition, and predicted microbial functions showed only minor differences between the fathers and the partners (Extended Data Fig. 1a, Extended Data Table 2a and Extended Fig. 2a-c) and there were no systematic differences between the microbiota profiles of the left and right axillae. Moreover, the left and right microbiota were closely correlated at the individual level, both in terms of alpha diversity (Pearson correlation: r = 0.796, p \u0026lt; 0.0001, for microbiota richness and 0.797 for Shannon diversity) and composition (Extended Data Fig. 1c,d). Therefore, for all subsequent analyses, we used individual-level microbiota profiles generated by aggregating the left and right microbiota for each individual.\u003c/p\u003e\n\u003cp\u003eTo test if there was a higher within-dyad similarity than expected by chance, we compared average microbiota dissimilarity between actual father-partner dyads with a distribution of average dissimilarities for randomly paired fathers and partners (n = 999 permutations). This approach provided consistently significant results for dissimilarities calculated based on amplicon sequence variant (ASV) prevalence and relative abundances. These results remained unchanged whether we analysed the whole dataset or only a subset of pairs with no missing sample (i.e., profiles of both left and right axilla of both father and partner, n = 49 dyads, Extended Data Table 2b) and were also significant if we analysed similarity in predicted microbiota functions (Extended Data Fig. 2c,d).\u003c/p\u003e\n\u003cp\u003eTo assess the potential impact of environmental bacteria on father-partner axillary microbiota similarity, environmental microbiota samples were also collected. The taxonomic content of environmental microbiota samples was clearly distinct from axillary microbiota (Extended Data Fig. 3a,b). As in the case of axillary microbiota, environmental samples of father-partner dyads exhibited higher similarity than randomly paired individuals (p = 0.019, SES = 2.24 for Bray-Curtis and p = 0.018, SES = 2.47 for Jaccard dissimilarities). According to differential abundance analysis, nine bacterial ASVs were overrepresented in the environmental samples (Extended Data Fig. 3c), representing 39.5% of high-quality sequences of control samples but only 1.8% of reads of the axillary microbiota profiles. Importantly, when these ASVs were excluded from the axillary microbiota profiles, we still detected higher similarity of axillary profiles than expected by chance (p = 0.008, SES = 2.53 for Bray-Curtis and p = 0.004, SES = 2.94 for Jaccard dissimilarities) (Extended Data Fig. 3d). However, similarity between partners\u0026rsquo; and fathers\u0026rsquo; environmental samples did not deviate from random expectation after the exclusion of ASVs overrepresented in environmental microbiota (p = 0.319, SES = 0.45 for Bray-Curtis and p = 0.179, SES = 0.91 for Jaccard dissimilarities) (Extended Data Fig. 3e). In conclusion, these analyses suggest that similarity in axillary microbiota of fathers and partners is not caused by the similarity in environmental bacterial pools.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eChemical compounds similarity\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn 41 father-partner dyads, we succeeded in analysing, in a single measurement sequence, the left and right axillary odour samples using comprehensive two-dimensional gas chromatography coupled with time-of-flight mass spectrometric detection. Even a visual inspection of the obtained chromatograms revealed, in some cases, a marked similarity of the chemical profiles between the father and the corresponding partner, as shown on selected chromatograms in Fig. 3a. The chromatograms were then processed using the recently developed tile-based alignment algorithm in the program ChromaTOF-Tile. Retrieval of more than 1,500 chemical features initially detected across all samples according to the inclusion criteria inferred from the pilot experiment (F-ratio = 2, S/N ratio = 50) followed by manual removal of artifact peaks resulted in the final set of 341 consistently occurring human sample-specific compounds. Their tentative identifications and peak volumes are listed in Supplementary Table Dataset 2. Among these compounds, we identified an expected set of human body odour analytes, especially homologous series of fatty acids, wax esters, hydrocarbons, aldehydes, ketones, amides, and steroids.\u003c/p\u003e\n\u003cp\u003eRelative abundances of square root-transformed peak volumes of the 341 compounds were treated as three separate datasets to decipher the chemical similarities among all participants for left, right and left-right averaged axillary samples. Fig. 3b shows a representation of the factor scores of the first three components of a principal component analysis (PCA) for left-right averaged samples. The distributions of father and partner scores largely overlap and do not indicate any consistent age-related separation between the two groups. The factor scores were then used to calculate the Euclidean distances among all pairs of samples, a proxy for chemical dissimilarity. These distances are plotted for fathers and corresponding partners in Fig. 3c. Comparison of the averaged within-dyad father-partner distance with the distribution of average distances for randomly paired fathers and partners (n = 999 permutations) revealed that the observed distances are significantly shorter than would be expected by chance (p = 0.001; Fig. 3d). Likewise, within-dyad father-partner distances were significantly shorter than the average distances of the corresponding fathers to all extra-pair partners (Wilcoxon Signed Rank Test, p = 0.0006; Fig. 3e).\u003c/p\u003e\n\u003cp\u003eWe obtained similar results when the same calculations were performed separately for left and right axillary samples (Extended Data Fig. 4). That is, average father-partner distance was significantly shorter than would be observed by chance (n = 999 permutations, p = 0.001 for both right and left axillary samples), and within-dyad father-partner distances were significantly shorter than average distances between individual fathers and all extra-pair partners (Wilcoxon Signed Rank Test, p = 0.006 for left and p \u0026lt; 0.0001 for right axillary samples). At the same time, the distances between all possible pairs of participants, including the observed father-partner distances, were significantly correlated between the right and the left axillary samples dataset (linear regression, r\u003csup\u003e2\u003c/sup\u003e = 0.473, p \u0026lt; 0.0001 for all pairs, r\u003csup\u003e2\u003c/sup\u003e = 0.391, p \u0026lt; 0.0001 for within-dyad father-partner distances; Extended Data Fig. 4f), as were also the distances between the left and the right axillary samples dataset, and the left-right averaged dataset (Extended Data Fig. 4i,j,k, respectively). The factor scores and variable loadings for the first three principal components and the resulting Euclidean distances between all possible pairs of participants are listed for the three analysed datasets in Supplementary Table Dataset 2.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRelationship quality with father\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePrevious studies indicate a positive association between reported quality of the relationship between daughter and father and the phenotypic similarity between her romantic partner and her father. Therefore, we also examined the perceived father-partner odour similarity and three measures of relationship quality. The first two measures concerned the daughter\u0026rsquo;s relationship quality with her father during childhood (assessed using the s-EMBU scale and by a single-item overall assessment of the relationship quality during childhood), while the third addressed the quality of their current relationship (using the ECR-RS scale). Correlations between these variables were generally moderate (Spearman\u0026apos;s rho \u0026lt; 0.4; Supplementary Table3). We found that perceived similarity was statistically significantly associated with the overall childhood relationship quality measure (Spearman\u0026rsquo;s rho = -0.28, p = 0.02), but not with any of the s-EMBU and ECR-RS subscales (Fig. 1e). None of the relationship quality measures showed significant association with either father-partner microbiome similarity (Extended Data Fig. 6) or chemical profile similarity (Extended Data Fig. 7). \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eOur results demonstrate the existence of an olfactory parental imprinting-like effect in humans which is consistently observed across connected but different levels of analysis. First, we show that the axillary odours of a woman\u0026rsquo;s father and romantic partner are perceptually more similar than expected by chance. Second, since axillary odour results from microbial activity at the skin surface, we also compared the composition of axillary microbiota in the same father-partners dyads. Using microbiota profiling based on 16S rRNA amplicon sequencing, we discovered significantly higher similarity in microbial diversity in these actual dyads than in randomly generated permutations. Third, by analysing the chemical profiles underlying the fathers\u0026rsquo; and partners\u0026rsquo; odours using GC\u0026times;GC-TOFMS, we found significantly higher similarity in father-partner dyads as compared to randomly generated pairs. \u003c/p\u003e\n\u003cp\u003eParental imprinting effects on adult mate choice have been demonstrated in cross-fostering studies in various mammals, from rodents\u003csup\u003e1\u003c/sup\u003e to ungulates\u003csup\u003e5\u003c/sup\u003e. While the nature of human research means that similar studies are correlational, similarity between a romantic partner and the opposite-sex parent has been reported in several other appearance-based characteristics, such as facial shape\u003csup\u003e13\u003c/sup\u003e or hair and eye colour\u003csup\u003e10\u003c/sup\u003e. The formation of such preferences may occur either continuously during childhood (perhaps, until adulthood) or be restricted to a specific sensitive juvenile period. In non-human mammals, olfaction-mediated social preferences form relatively early in ontogeny\u003csup\u003e16\u003c/sup\u003e but might be modulated by subsequent experience\u003csup\u003e17\u003c/sup\u003e. In humans, this remains poorly understood\u003csup\u003e12\u003c/sup\u003e. Earlier studies suggested that 1-5 years might be the most sensitive period\u003csup\u003e19,20\u003c/sup\u003e, while more recent investigation points to the early teenage years\u003csup\u003e21\u003c/sup\u003e. \u003c/p\u003e\n\u003cp\u003eWe suggest that similarity between a woman\u0026rsquo;s father and romantic partner results from a process of social learning of parental characteristics during childhood. A major assumption of this is that such characteristics are stable over a long period of time (i.e., between one\u0026rsquo;s childhood and young adulthood), as is known to occur in human faces\u003csup\u003e22,23\u003c/sup\u003e. While no similar data is available for body odours, studies on individual odour recognition\u003csup\u003e24\u003c/sup\u003e and genetic influences\u003csup\u003e25\u003c/sup\u003e do suggest such stability.\u003c/p\u003e\n\u003cp\u003eAn alternative explanation for the reported similarity between parent and romantic partner is that it arises as a by-product of self-similarity preferences\u003csup\u003e26\u003c/sup\u003e\u003cstrong\u003e.\u003c/strong\u003e Body odour of parents and children\u003csup\u003e27\u003c/sup\u003e or siblings\u003csup\u003e25\u003c/sup\u003e are perceived as more similar compared to non-related individuals. While we cannot rule this out, previous studies indicate that parental effects are stronger than self-similarity preferences\u003csup\u003e28\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eA further alternative is that perceived odour similarity of fathers and partners, and the underpinning similarity in chemical profiles that we report here, arises not due to the hypothesised imprinting-like effects, but rather as a result of the daughter\u0026rsquo;s association with the two men. For example, learned dietary preferences and lifestyle characteristics in the natal household could be replicated in the woman\u0026rsquo;s adult home that she shares with the partner. If these behavioural patterns shape the microbial populations present in the natal and adult homes, they could in turn generate similarities in the chemical profiles of those living there, and of course in their emergent odour. However, analysis of the microbiome in environmental samples collected at the homes of the participants, demonstrated that the reported pattern cannot be explained simply by similarities in their environment. \u003c/p\u003e\n\u003cp\u003eTheoretical models predict that mate preferences should be flexible to current environmental fluctuations\u003csup\u003e29\u003c/sup\u003e. Why, then, would selection shape adult mate preferences according to parental cues so early in ontogeny? The answer may be that complete flexibility is demanding to maintain or potentially disadvantageous in a stable environment. Using\u003cstrong\u003e \u003c/strong\u003eparental characteristics as a template for own mate choice might be a mechanism for outbreeding avoidance or preference for time-tested and locally adapted phenotypes\u003cstrong\u003e. \u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFinally, some studies have indicated positive associations between the level of similarity between the opposite-sex parent and the offspring\u0026rsquo;s romantic partner and the quality of the relationship between parent and offspring\u003csup\u003e13,14\u003c/sup\u003e. Although its robustness should be confirmed in future studies, our results partly support such an association between perceived odour similarity and relationship quality. The fact that relationship quality was, in contrast, not associated with either microbial or chemical profiles, suggests that the learned odour which appears to influence mate preference is formed from a subset of the entire chemical profile. This is to be expected given the chemical diversity of axillary odour and the multiple forms of social information it appears to carry\u003csup\u003e30\u003c/sup\u003e. Altogether, our findings reveal a chain of mechanisms that influence human social odour preferences, and open new directions in how early family influences may impact on social functioning in adulthood and in romantic relationships in particular.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data that support the findings of the current study are available at Supplementary Table Dataset 1 and Dataset 2.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe study was supported by a Czech Science Foundation grant (no.18-15168S). R.H., R.B. and P.K. were supported by the Institute of Organic Chemistry and Biochemistry, Czech Academy of Sciences (RVO 61388963). We are indebted to Krist\u0026yacute;na Dachsov\u0026aacute;, Anna Fi\u0026scaron;erov\u0026aacute;, Krist\u0026yacute;na Moln\u0026aacute;rov\u0026aacute;, Žaneta P\u0026aacute;tkov\u0026aacute;, Kateřina Roberts, and Krist\u0026yacute;na \u0026Scaron;\u0026iacute;pkov\u0026aacute; for their help with data collection, Vladim\u0026iacute;r Kunc for programming the rating application and technical support, Pavel \u0026Scaron;ebesta and V\u0026iacute;t Třebick\u0026yacute; for their help with the manuals, and to all the participants for their cooperation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eConceptualization: L.J., Z.\u0026Scaron;., R.H., J.K., and J.H.; methodology: L.J., Z.\u0026Scaron;. R.H., J.K., P.K., J.T.F., and J.H.; project administration: L.J., and Z.\u0026Scaron;.; data collection: L.J., Z.\u0026Scaron;.,D.S., and J.T.F.; chemical analysis: R.H., P.B., and R.B.; microbiome analysis: J.K., and L.S.; statistical analysis: L.J., R.H., and J.K.; visualization: L.J., R.H., J.K., and D.S.; funding acquisition: J.H.; writing of the original draft: L.J., R.H., J.K., and J.H.; writing, review and editing: L.J., Z.\u0026Scaron;., R.H., J.K., P.K., R.B., L.S., J.T.F., D.S., S.C.R., and J.H.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eReprints and permissions information\u003c/strong\u003e is available at www.nature.com/reprints.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003ePenn, D. \u0026amp; Potts, W. MHC-disassortative mating preferences reversed by cross-fostering. \u003cem\u003eProc. R. Soc. London Ser. B-Biological Sci.\u003c/em\u003e \u003cstrong\u003e265\u003c/strong\u003e, 1299\u0026ndash;1306 (1998).\u003c/li\u003e\n\u003cli\u003eBeauchamp, G. K. \u0026amp; Wellington, J. L. Cross-species rearing influences urine preferences in wild guinea pigs. \u003cem\u003ePhysiol. Behav.\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 1121\u0026ndash;1124 (1981).\u003c/li\u003e\n\u003cli\u003eMcGann, P. John. Poor human olfaction is a 19th-century myth. \u003cem\u003eScience (80-. ).\u003c/em\u003e \u003cstrong\u003e356\u003c/strong\u003e, 6338 (2017).\u003c/li\u003e\n\u003cli\u003eRavreby, I., Snitz, K. \u0026amp; Sobel, N. There is chemistry in social chemistry. \u003cem\u003eSci. Adv.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, (2022).\u003c/li\u003e\n\u003cli\u003eKendrick, K. M., Hinton, M. R., Atkins, K., Haupt, M. A. \u0026amp; Skinner, J. D. Mothers determine sexual preferences. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e395\u003c/strong\u003e, 229\u0026ndash;230 (1998).\u003c/li\u003e\n\u003cli\u003eTheis, K. R. \u003cem\u003eet al.\u003c/em\u003e Symbiotic bacteria appear to mediate hyena social odors. \u003cem\u003eProc. Natl. Acad. Sci. U. S. A.\u003c/em\u003e \u003cstrong\u003e110\u003c/strong\u003e, 19832\u0026ndash;19837 (2013).\u003c/li\u003e\n\u003cli\u003eZomer, S. \u003cem\u003eet al.\u003c/em\u003e Consensus multivariate methods in gas chromatography mass spectrometry and denaturing gradient gel electrophoresis: MHC-congenic and other strains of mice can be classified according to the profiles of volatiles and microflora in their scent-marks. \u003cem\u003eAnalyst\u003c/em\u003e \u003cstrong\u003e134\u003c/strong\u003e, 114\u0026ndash;123 (2009).\u003c/li\u003e\n\u003cli\u003eArchie, E. A. \u0026amp; Theis, K. R. Animal behaviour meets microbial ecology. \u003cem\u003eAnim. Behav.\u003c/em\u003e \u003cstrong\u003e82\u003c/strong\u003e, 425\u0026ndash;436 (2011).\u003c/li\u003e\n\u003cli\u003eBoothroyd, L. G. \u0026amp; Vukovic, J. Mate Preferences Across the Lifespan. in \u003cem\u003eThe Oxford Handbook of Evolutionary Psychology and Behavioral Endocrinology\u003c/em\u003e (eds. Welling, L. L. M. \u0026amp; Schackelford, T. K.) (Oxford University Press, 2018).\u003c/li\u003e\n\u003cli\u003eLittle, A. C., Penton-Voak, I. S., Burt, D. M. \u0026amp; Perrett, D. I. Investigating an imprinting-like phenomenon in humans Partners and opposite-sex parents have similar hair and eye colour. \u003cem\u003eEvol. Hum. Behav.\u003c/em\u003e \u003cstrong\u003e24\u003c/strong\u003e, 43\u0026ndash;51 (2003).\u003c/li\u003e\n\u003cli\u003ePerrett, D. I. \u003cem\u003eet al.\u003c/em\u003e Facial attractiveness judgements reflect learning of parental age characteristics. \u003cem\u003eProc. R. Soc. London Ser. B-Biological Sci.\u003c/em\u003e \u003cstrong\u003e269\u003c/strong\u003e, 873\u0026ndash;880 (2002).\u003c/li\u003e\n\u003cli\u003eAkao, K. A., Adair, L. \u0026amp; Brase, G. L. Parental attachment style, but not environmental quality, is associated with use of opposite-sex parents as a template for relationship partners. \u003cem\u003ePers. Individ. Dif.\u003c/em\u003e (2016) doi:10.1016/j.paid.2016.04.071.\u003c/li\u003e\n\u003cli\u003eWiszewska, A., Pawlowski, B. \u0026amp; Boothroyd, L. G. Father-daughter relationship as a moderator of sexual imprinting: a facialmetric study. \u003cem\u003eEvol. Hum. Behav.\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 248\u0026ndash;252 (2007).\u003c/li\u003e\n\u003cli\u003eBereczkei, T., Gyuris, P. \u0026amp; Weisfeld, G. E. Sexual imprinting in human mate choice. \u003cem\u003eProc. R. Soc. B\u003c/em\u003e \u003cstrong\u003e271\u003c/strong\u003e, 1129\u0026ndash;1134 (2004).\u003c/li\u003e\n\u003cli\u003eJacob, S., McClintock, M. K., Zelano, B. \u0026amp; Ober, C. Paternally inherited HLA alleles are associated with women\u0026rsquo;s choice of male odor. \u003cem\u003eNat. Genet.\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 175\u0026ndash;179 (2002).\u003c/li\u003e\n\u003cli\u003ePorter, R. H. \u0026amp; Etscorn, F. Olfactory imprinting resulting from brief exposure in Acomys cahirinus. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e250\u003c/strong\u003e, 732\u0026ndash;733 (1974).\u003c/li\u003e\n\u003cli\u003eNyby, J., Whitney, G., Schmitz, S. \u0026amp; Dizinno, G. Postpubertal experience establishes signal value of mammalian sex odor. \u003cem\u003eBehav. Biol.\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 545\u0026ndash;552 (1978).\u003c/li\u003e\n\u003cli\u003eBoothroyd, L. G. \u0026amp; Vukovic, J. Mate Preferences Across the Lifespan. in \u003cem\u003eThe Oxford Handbook of Evolutionary Psychology and Behavioral Endocrinology\u003c/em\u003e (eds. Welling, L. W. \u0026amp; Shackelford, T. K.) (OUP, 2018).\u003c/li\u003e\n\u003cli\u003eEnquist, M., Aronsson, H., Ghirlanda, S., Jansson, L. \u0026amp; Jannini, E. A. Exposure to Mother\u0026rsquo;s Pregnancy and Lactation in Infancy is Associated with Sexual Attraction to Pregnancy and Lactation in Adulthood. \u003cem\u003eJ. Sex. Med.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 140\u0026ndash;147 (2011).\u003c/li\u003e\n\u003cli\u003eSaxton, T. K. Experiences during specific developmental stages influence face preferences. \u003cem\u003eEvol. Hum. Behav.\u003c/em\u003e \u003cstrong\u003e37\u003c/strong\u003e, 21\u0026ndash;28 (2016).\u003c/li\u003e\n\u003cli\u003e\u0026Scaron;těrbov\u0026aacute;, Z., Tureček, P., Havl\u0026iacute;ček, J. \u0026amp; Varella Valentova, J. \u003cem\u003eWomen learn paternal characteristics used in mate choice during early childhood and adolescence\u003c/em\u003e. doi:10.31234/osf.io/fhrp8.\u003c/li\u003e\n\u003cli\u003eBulygina, E., Mitteroecker, P. \u0026amp; Aiello, L. Ontogeny of facial dimorphism and patterns of individual development within one human population. \u003cem\u003eAm. J. Phys. Anthropol.\u003c/em\u003e \u003cstrong\u003e131\u003c/strong\u003e, 432\u0026ndash;443 (2006).\u003c/li\u003e\n\u003cli\u003eMileva, M., Young, A. W., Jenkins, R. \u0026amp; Burton, A. M. Facial identity across the lifespan. \u003cem\u003eCogn. Psychol.\u003c/em\u003e \u003cstrong\u003e116\u003c/strong\u003e, 101260 (2020).\u003c/li\u003e\n\u003cli\u003eHold, B. \u0026amp; Schleidt, M. The Importance of Human Odour in Non‐verbal Communication. \u003cem\u003eZ. Tierpsychol.\u003c/em\u003e \u003cstrong\u003e43\u003c/strong\u003e, 225\u0026ndash;238 (1977).\u003c/li\u003e\n\u003cli\u003eRoberts, S. C. \u003cem\u003eet al.\u003c/em\u003e Body Odor Similarity in Noncohabiting Twins. \u003cem\u003eChem. Senses\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 651\u0026ndash;656 (2005).\u003c/li\u003e\n\u003cli\u003eAllen, C., Havl\u0026iacute;ček, J., Williams, K. \u0026amp; Roberts, S. C. Evidence for odour-mediated assortative mating in humans: The impact of hormonal contraception and artificial fragrances. \u003cem\u003ePhysiol. Behav.\u003c/em\u003e \u003cstrong\u003e210\u003c/strong\u003e, 112541 (2019).\u003c/li\u003e\n\u003cli\u003ePorter, R. H., Cernoch, J. M. \u0026amp; Balogh, R. D. Odor signatures and kin recognition. \u003cem\u003ePhysiol Behav\u003c/em\u003e \u003cstrong\u003e34\u003c/strong\u003e, 445\u0026ndash;448 (1985).\u003c/li\u003e\n\u003cli\u003e\u0026Scaron;těrbov\u0026aacute;, Z., Tureček, P. \u0026amp; Kleisner, K. Consistency of mate choice in eye and hair colour: Testing possible mechanisms. \u003cem\u003eEvol. Hum. Behav.\u003c/em\u003e \u003cstrong\u003e40\u003c/strong\u003e, 74\u0026ndash;81 (2019).\u003c/li\u003e\n\u003cli\u003eAh-King, M. \u0026amp; Gowaty, P. A. A conceptual review of mate choice: stochastic demography, within-sex phenotypic plasticity, and individual flexibility. \u003cem\u003eEcol. Evol.\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 4607\u0026ndash;4642 (2016).\u003c/li\u003e\n\u003cli\u003eHavl\u0026iacute;ček, J., Fialov\u0026aacute;, J. \u0026amp; Roberts, S. C. Individual Variation in Body Odor. in \u003cem\u003eSpringer Handbook of OdorHandbook of o\u003c/em\u003e (ed. Buettner, A.) 949\u0026ndash;961 (Springer, 2017).\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eParticipants\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe sample consisted of 67 triads: heterosexual women (mean age = 24.2 years, SD = 3.3), their fathers (mean age = 51.5 years, SD = 6.3) and romantic partners (mean age = 26.7 years, SD = 4.4). The required sample size was estimated based on a priori power analysis in which, based on previous studies, we assumed an association between partner\u0026rsquo;s and opposite-sex parent\u0026rsquo;s characteristics to be equivalent to a positive correlation of value 0.25 (for details, see the study\u0026rsquo;s preregistration at https://osf.io/adrtc). Only women who met the following criteria were enrolled in the study: i) aged 18-35 years, ii) currently in a committed long-term (for a minimum of 6 months) heterosexual romantic relationship (actual mean length = 59 months, SD = 45.3, range = 8-233); iii) shared a common household with their biological father until at least 8 years of age (actual mean = 20.9, SD = 3.6, range = 8-30 years); and iv) not suffering from conditions affecting their sense of smell which may in turn influence their olfactory preferences\u003csup\u003e31\u003c/sup\u003e. The enrolment criteria for fathers and romantic partners were: i) aged under 65 years (fathers) or 18-40 years (partners) ii) non-smokers; iii) in good health and having no underlying condition which affects body odour (e.g., diabetes)\u003csup\u003e30\u003c/sup\u003e; iv) not currently using medication; and v) having unshaven axillae\u003csup\u003e32\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eParticipants were recruited via an e-mail contact list compiled from our previous studies, social media (Facebook, Instagram), and flyers advertised at Charles University, Prague. To check whether participants met the enrolment criteria (e.g., age, health conditions), they completed an online contact form via the Qualtrics platform. Subsequently, we arranged a personal meeting with one individual from each triad and handed over the package with material for data collection. During this meeting, participants were informed about the aims of the study and the sampling procedure was fully explained by going over the written instructions which accompanied the package.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eProcedure\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWithin the space of one week, each woman completed questionnaires to assess the quality of the relationship with her father during childhood (s-EMBU and overall assessment of childhood relationship quality) and her current attachment style (ECR-RS). Fathers and partners were asked to provide body odour samples for perceptual ratings, chemical analysis, and axillary microbiome analysis. Participants also provided facial photographs and voice recordings which are not the subject of this paper and will be reported elsewhere.\u003c/p\u003e\n\u003cp\u003eData collection was carried all over the Czech Republic (32 villages, towns or cities, Fig. 1a), increasing the variability of socio-demographic characteristics of our sample (Extended data Table 1). Participants confirmed their participation by signing a consent form. Each triad received compensation for their participation in the sum of 1200 CZK (app. 44 EUR). The study was preregistered (https://osf.io/adrtc) and was approved by the Institutional Review Board of Charles University, Faculty of Science (no. 2017/20).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eBody odour collection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eEach father and partner obtained a package including printed instructions and materials for sample collection (non-perfumed liquid soap, tissues, six sterile cotton swabs, three glass vials, three Eppendorf tubes, a vial with the sterile physiological solution, three cotton pads separately enclosed in the labelled zip-lock plastic bags, surgical tape, white cotton T-shirt and a polystyrene transport box (15x20x10 cm). Participants were asked to refrain from consuming aromatic food (a list was provided), drinking alcohol, smoking or using drugs, using cosmetics (e.g., perfumes and deodorants) and from strenuous physical activities including sex during the 48 hours preceding sample collection. All these activities may influence body odour quality\u003csup\u003e33\u003c/sup\u003e. After collection of the samples, each participant completed a survey to document rule compliance (Supplementary Table 1). Although there were some minor violations two days before sampling, there were none reported in the 24h period leading up to the sampling.\u003c/p\u003e\n\u003cp\u003eIn the evening on the sampling day, participants collected axillary samples for chemical analysis and then for microbiome analysis. They first washed their hands using non-perfumed liquid soap and dried them with the tissues. To collect samples for chemical analysis, participants wiped each axilla 20 times with a clean cotton swab and placed each swab in a 2 ml empty glass vial. The same procedure was used for collecting axillary microbiome samples, except that the swabs were first soaked in sterile physiological solution. In both types of sampling, they used a sterile cotton swab, unique for each axilla. Axillary microbiome swabs were placed inside an Eppendorf vial containing absolute (99.8%) ethanol. Finally, two control samples (one each for chemical and axillary microbiome analysis) were collected by waving a clean cotton swab 20 times in the air); these samples were used to assess possible effect of environment (e.g., specific odour of the household). All these samples were put into the pre-labelled zip-lock plastic bags and odourless polystyrene box and placed in a freezer (-20\u0026deg;C) immediately after sampling (Fig. 1b).\u003c/p\u003e\n\u003cp\u003eSubsequently, the participants took a shower and washed with the provided non-perfumed liquid soap. They then attached a cotton pad to each axilla using surgical tape. To minimise odour contamination from clothing and other extrinsic ambient odours, participants wore a new white 100% cotton T-shirt as the first layer of their clothing. They wore the pads for the following 12 hours (overnight)\u003csup\u003e34,35\u003c/sup\u003e. Participants collected a further control sample by placing a clean cotton pad on their bedside table for the same period that they wore the axillary pads. Next morning, they placed the cotton pads in separate, pre-labelled zip-lock plastic bags and added these to the polystyrene box together with the samples for chemical and axillary microbiome analysis, which was then put back into the freezer\u003csup\u003e36\u003c/sup\u003e. They returned all samples to the experimenters within a week (in most cases, the experimenters collected these from the participant\u0026rsquo;s homes). For an overview of the entire procedure, see Supplementary Fig. 1.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eQuestionnaires \u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe quality of relationship with the father during childhood was assessed using the s-EMBU scale (an acronym for \u0026lsquo;Egna Minnen Betr\u0026auml;ffande Uppofostran\u0026rsquo;, Swedish for \u0026lsquo;My memories of upbringing\u0026rsquo;)\u003csup\u003e37\u003c/sup\u003e. This scale is a widely used tool for assessing adults\u0026rsquo; perception of their parents\u0026rsquo; rearing behaviour. It consists of 23 items, loading onto three sub-scales: Rejection (7 items; possible range 7\u0026ndash;28), Emotional Warmth (7 items; possible range 7\u0026ndash;28), and (Over)Protection (9 items; possible range 9\u0026ndash;36). Responses were indicated on a 4-point Likert-type scale (ranging from 1 \u0026ndash; \u0026lsquo;No, never\u0026rsquo; to 4 \u0026ndash; \u0026lsquo;Yes, most of the time\u0026rsquo;). We used a Czech language version used in previous studies\u003csup\u003e38\u003c/sup\u003e. We further used a single item to explore overall quality of the relationship during childhood (\u0026lsquo;How would you evaluate your relationship with your father during childhood?\u0026rsquo;), indicated on a 7-point Likert-type scale (ranging from 1 \u0026ndash; \u0026lsquo;Very negative\u0026rsquo; to 7 \u0026ndash; \u0026lsquo;Very positive\u0026rsquo;).\u003c/p\u003e\n\u003cp\u003eAdult attachment style between the woman and her father was assessed using the Experiences in Close Relationships \u0026ndash; Relationship Structures (ECR-RS) questionnaire\u003csup\u003e39\u003c/sup\u003e. This consists of 9 items about attachment orientation of the daughter to father and comprises two subscales: Anxiety (3 items; possible range 3\u0026ndash;21) and Avoidance (6 items; possible range 3\u0026ndash;42). Daughters completed a Czech version of the scale using a 7-point Likert-type scale (ranging from 1 \u0026ndash; \u0026lsquo;Strongly disagree\u0026rsquo; to 7 \u0026ndash; \u0026lsquo;Strongly agree\u0026rsquo;). Mean scores for each questionnaire are reported in Extended data Table 1.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePerceptual Ratings\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRaters\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn total, 306 female raters (mean age = 24.0, SD = 4.0) took part in the rating sessions. They were recruited via an e-mail contact list compiled from our previous studies, social media (Facebook, Instagram) and flyers or by oral invitation around Charles University and the Czech Technical University. Only women in good respiratory health and without any olfactory malfunction took part in the rating sessions. Each rater could participate only in a single rating session. At the end of the session, raters were informed about the study goals and obtained 150 CZK (approximately 5.5 EUR) as compensation for their time.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMatch-to-sample tests \u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo assess perceived similarity between father and partner body odour we used a match-to-sample paradigm. In total, 67 body odour sets were created. Each set consisted of a father\u0026rsquo;s body odour sample (the \u0026lsquo;target\u0026rsquo;), the partner\u0026rsquo;s body odour sample (the \u0026lsquo;match\u0026rsquo;) and body odour samples from three other individuals (\u0026lsquo;distractors\u0026rsquo;). The distractor body odours were collected from independent male donors (N= 89, mean age = 25.8; SD = 5.4) who met the same conditions as partners (e.g., the same age, being non-smokers, having unshaven axillae) and who followed the same body odour sampling procedure as described above. Each set consisted only of samples collected from one axillary side (i.e., left/right). In total, 28 sets consisted of left axillary samples and 39 sets used right axillary samples. We sometimes also used partners\u0026rsquo; body odour samples as distractors, but these were always the other axillary sample (in other words, if a sample from the left axilla was used as a match, the right axilla sample was used as a distractor).\u003c/p\u003e\n\u003cp\u003eIn each session (13 sessions in total), 5 sets of stimuli were used (except the last two sessions consisting of 6 sets). The initial position of the match and the distractors was randomized for each rater (Fig. 1c). The match sample position was then rotated clockwise for each subsequent rater. The distractors used for each rater were randomly chosen from the pool of 15 (or 18 in the last two sessions) distractors available for the given rating session. Similarly, the order of the individual sets was randomized for each rater. Samples were randomized by a blinded experimenter who followed a randomization matrix generated before the rating session.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRating session\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eRating sessions took place in a quiet, ventilated room with a mean temperature of 22.5\u0026deg;C. The body odour samples were thawed one hour before the rating session\u003csup\u003e36\u003c/sup\u003e. Subsequently, the samples were enclosed in 250ml opaque labelled jars. Raters provided informed consent and were informed about the rating task. The origin of the body odour samples was not disclosed.\u003c/p\u003e\n\u003cp\u003eFirst, raters were asked to evaluate similarity of the body odours within each set (\u0026lsquo;Please sniff the target and then the four other samples. Order the four samples from the most similar to the least similar as compared to the target\u0026rsquo;). Raters marked their scores into tablet using a purpose-built application MatchApp. The sniffing time was not restricted, most raters finished the odour rating task in 30 minutes. The body odour ratings were followed by assessments of facial photographs and voice recordings, results of which are not relevant to this study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical analysis of perceptual ratings\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe perceived father-partner similarity in body odour was first evaluated using Chi-square tests - the perceived similarity rankings for each father-partner dyad were compared to the null hypothesis of uniform random distribution of the ranks (i.e., that the match (partner) should appear as the most similar of the four samples in 25 % of cases). Next, to analyse the relative frequency of the match appearing in each of the four ranks, while accounting for non-independence of individual rankings and target\u0026rsquo;s identity we employed an ordinal probit bayesian mixed model (fitted using R package brms)\u003csup\u003e40\u003c/sup\u003e where parameter estimates and 95% posterior credible intervals were reported. The identity of the raters, the identity of the targets, and axillary side (left/right) were all included as random effects. For a subset of the partner samples, there was no difference between left and right axilla in similarity ratings (estimate [\u0026plusmn; 95% posterior credibility intervals] = 0.04 \u0026plusmn; [-0.43~ 0.50]).\u003c/p\u003e\n\u003cp\u003eThe potential effect of relationship quality with the father on perceived father-partner odour similarity was analysed by non-parametric paired Spearman\u0026rsquo;s correlation (Extended data Table 2). Correlations between individual measures of the relationship quality were generally moderate (Spearman\u0026apos;s rho \u0026lt; 0.4) except for the strong negative association between the variables Overall relationship quality and EMBU Rejection (Spearman\u0026apos;s rho = -0.59) and the positive association of Overall relationship quality and EMBU Emotional Warmth (Spearman\u0026apos;s rho = 0.57) (Supplementary Table 3).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAxillary microbiome analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMetagenomic DNA was extracted using PowerSoil (Qiagen) kits and sequencing libraries were prepared using a two-step PCR approach, following Glenn et al.\u003csup\u003e41\u003c/sup\u003e. Gene-specific primers covering the V3-V4 variable region of bacterial 16S rRNA (i.e., S-D-Bact-0341-b-S-17 [CCTACGGGNGGCWGCAG] and S-D-Bact-0785-a-A-21 [GACTACHVGGGTATCTAATCC]) were used during the first PCR step\u003csup\u003e42\u003c/sup\u003e. The primers were extended at the 5\u0026rsquo; end with \u0026ldquo;tails\u0026rdquo; serving as a priming site for the second PCR (i.e., the first 33 bp and 34 bp of read1 and read2 Nextera adapters for F and R primers, respectively). The first PCR was performed in 10 \u0026mu;l and consisted of 5 \u0026mu;l 1x KAPA HiFi Hot Start Ready Mix (Roche), each primer at 0.2 \u0026mu;M and 4.6 \u0026mu;l DNA. PCR conditions were as follows: initial denaturation at 95\u0026deg;C for 3 min followed by 35 cycles each of 95\u0026deg;C (30 s), 55\u0026deg;C (30 s), and 72\u0026deg;C (30 s), and a final extension at 72\u0026deg;C (5 min). Dual-indexed Nextera sequencing adapters were reconstructed during the second PCR, which followed the first PCR conditions except that: it was performed in 20 \u0026mu;l volume; the concentration of each primer was 1 uM; 6\u0026mu;l of the first PCR product were used as a template; and the number of PCR cycles was 12. Technical duplicates were prepared to account for noise due to PCR and sequencing stochasticity.\u003c/p\u003e\n\u003cp\u003eProducts of the second PCR were run on 1.5% agarose gel and their concentrations were assessed based on gel band intensities using GenoSoft software (VWR International, Belgium). They were pooled equimolarly and size-selected with Pippin Prep (Sage Science, USA) at 520 - 750 bp. Resulting libraries were sequenced using MiSeq (Illumina, USA) and v3 chemistry (i.e., 2 \u0026times; 300 bp paired-end reads). Sequencing data are available at the European Nucleotide Archive under the accession number of the whole project (\u003cem\u003ewill be completed upon the acceptance of the manuscript\u003c/em\u003e).\u003c/p\u003e\n\u003cp\u003eSamples were demultiplexed and primers were trimmed by \u003cem\u003eskewer\u003c/em\u003e software\u003csup\u003e43\u003c/sup\u003e. Using \u003cem\u003edada2\u003c/em\u003e\u003csup\u003e44\u003c/sup\u003e, we eliminated low-quality sequences (expected number of errors \u0026gt; 2), denoised quality-filtered fastq files, and constructed abundance matrix representing read counts for individual amplicon sequencing variants (hereafter ASV) in each sample. Next, we identified chimeric ASVs using \u003cem\u003euchime\u003c/em\u003e\u003csup\u003e45\u003c/sup\u003e and gold.fna reference database and eliminated them from the abundance matrix. Taxonomic assignation of ASVs was conducted by \u003cem\u003eRDP\u003c/em\u003e \u003cem\u003eclassifier\u003c/em\u003e (80% confidence threshold\u003csup\u003e46\u003c/sup\u003e) and Silva database (v. 132)\u003csup\u003e47\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eTechnical replicates of axillary microbiota samples exhibited high consistency in both composition (Procrustean correlation, r = 0.99, p = 0.0001) and Shannon diversity (Pearson\u0026apos;s r = 0.93, p \u0026lt; 0.0001). Consequently, we merged sequences corresponding to individual samples for all subsequent analysis. To avoid microbial diversity inflation due to PCR and sequencing artifacts, we eliminated all OTUs that were present just once within the set of PCR replicates for a given sample\u003csup\u003e48\u003c/sup\u003e. Prior to statistical calculations, we also excluded one sample that exhibited low consistency among its technical replicates and one sample of low sequencing coverage (\u0026lt; 1500 high quality sequences). The final dataset included 246 samples of left and/or right axillary microbiota coming from 132 individuals (66 fathers and 66 partners). These samples were represented by 3,129,530 high-quality reads that were distributed among 4,891 ASVs. The median sequencing coverage was 13,152 reads per sample (range = 1,750 \u0026ndash; 34,343). Functional predictions of bacterial metagenome were conducted by PICRUSt2 pipeline\u003csup\u003e49\u003c/sup\u003e using the default setup. Predicted metagenomes were then categorized to functional pathway\u003csup\u003e50\u003c/sup\u003e and their predicted abundances in each sample were used in later statistical analyses. Weighted Nearest Sequenced Taxon Index (weighted NSTI) was 0.02755007 +/- 0.001274799 (+/- S.E) suggesting very high precision of metagenome predictions.\u003c/p\u003e\n\u003cp\u003eIn addition to the axillary microbiota samples, we also collected 74 environmental microbiota samples. Due to lower DNA concentrations and PCR stochasticity, duplicates from the environmental samples had lower alpha diversity (r = 0.74, p \u0026lt; 0.0001) and compositional consistency (r = 0.68, p \u0026lt; 0.0001) compared to duplicates from the axillary microbiota, and often a lower number of reads. Therefore, we focused our analyses on 50 ambient environmental microbiota samples that had \u0026gt;1000 reads after all filtering steps.\u003c/p\u003e\n\u003cp\u003eMicrobial alpha diversity did not significantly vary between left and right armpit (Generalized Linear Mixed Model (GLMM): \u0026Delta; D.f. = 1, \u0026chi;2= 1.31, p = 0.252 for ASVs richness and \u0026Delta; D.f. = 1, \u0026chi;2= 1.31, p = 0.252 for Shannon diversity) and there was also no significant difference in the number of detected ASVs between fathers and partners (GLMM: \u0026Delta; D.f. = 1 \u0026chi;2= 2.06, p = 0.151). However, partners exhibited slightly lower Shannon diversity compared to fathers (GLMM: estimate = -0.24 +/- 0.12, \u0026Delta; D.f. = 1, \u0026chi;2= 4.27, p = 0.039). NMDS (Extended Data Fig. 1a) and PERMANOVA analyses did not detect any systematic difference in the composition between left and right axilla, yet they revealed a slight compositional shift between fathers and partners explaining only \u0026lt; 2% of total variation in composition (Extended Data Table 3a and Extended Data Fig. 1b). Similarly, alpha diversity was closely correlated in the left and right axilla (Extended Data Fig 1c). According to GLMM analyses with distances to group-specific centroids as a response, profiles of left and right axilla were equally dispersed, but fathers exhibited higher dispersion in composition (i.e., inter-individual variation) than partners (Extended Data Table 3b and Extended Data Fig. 1d).\u003c/p\u003e\n\u003cp\u003eSamples from both left and right axillae were successfully profiled in the case of 114 individuals (58 fathers and 56 partners). Based on this subset, there was a tight correlation in alpha diversity between left and right axillae (Pearson\u0026rsquo;s correlation: r = 0.796, p \u0026lt; 0.0001 for ASVs richness and 0.797 for Shannon diversity). The same holds for consistency in the composition, where microbiota divergences calculated using ASVs presence/absence was more decisive (Procrustean correlation for Jaccard distances: r = 0.94, p = 0.0001) than divergence metrics utilizing relative ASVs abundances (Procrustean correlation for Bray-Curtis distances: r = 0.82, p = 0.00001, Fig. 2b and Extended Data Fig. 1c). Furthermore, consistency between left and right axillae was significantly higher for fathers than for partners (GLMM with Procrustean residuals as a response: \u0026Delta; D.f. = 1, \u0026chi;2= 5.44, p = 0.02 for Bray-Curtis and \u0026Delta; D.f. = 1, \u0026chi;2= 9.09, p = 0.003 for Jaccard dissimilarities, Extended Data Fig. 1d). For the subsequent analyses, profiles of left and right axillae of the same individual were aggregated.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical analysis of microbiome data\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eShannon indices and number of ASVs in each sample (hereafter ASVs richness) were calculated after dataset normalization by rarefaction (threshold = 1,750 reads per sample) and used as alpha diversity measures. We analysed differences in alpha diversity between fathers and partners and left vs. right axillae using GLMM with Gaussian error distribution. Individual identity nested within pair identity were specified as random effects. Pearson correlation was used to check for a correlation in alpha diversity between left and right axillae.\u003c/p\u003e\n\u003cp\u003eDifferences in composition of microbiota profiles was analysed based on Bray Curtis dissimilarities accounting for ASV relative abundances and binary Jaccard dissimilarities accounting just for ASVs presence/absence. Both these metrices were calculated after the rarefaction of the ASV abundance matrix. First, we checked if there were any differences in composition between left and right axillae and between fathers and partners using Nonmetric Multidimensional Scaling (NMDS) and PERMANOVA with permutation constraint (i.e., \u0026lsquo;strata\u0026rsquo;) setup to within-pair level. To test if interindividual variation differed between left and right axillae, and between fathers and partners, we exacted distance to centroids for father\u0026rsquo;s and partner\u0026rsquo;s samples and used them after log transformation as a response in GLMM, assuming Gaussian errors and the same random structure as described above. ASVs overrepresented in fathers or partners were identified by differential abundance analyses, where read counts for each ASV were included as a response into GLMM with negative binomial errors and log-scaled total sequencing depth for a given sample as an offset. The model random structure remained unchanged. To achieve model convergence, these models were fitted using glmmTMB package\u003csup\u003e51\u003c/sup\u003e only for ASVs detected in at least 10 samples. The Qvalue method\u003csup\u003e52\u003c/sup\u003e was then used to control for false discoveries due to multiple testing, with the threshold being setup to q \u0026lt; 0.05.\u003c/p\u003e\n\u003cp\u003eThe composition and alpha diversity of left and right axillary microbiota were tightly correlated at the individual level. We therefore aggregated the left and right axillary profiles for each individual and used the aggregated profiles to test if there was higher similarity between father-partner pairs than expected by chance. To do this, we compared average microbiota dissimilarity between actual father-partner dyads with a distribution of average dissimilarities drawn from randomly paired fathers and partners (n = 999 permutations). The outcome of these analyses were also expressed as Standardised Effect Sizes (SES)\u003csup\u003e53\u003c/sup\u003e using the formula SES = (\u0026sum;DIF\u003csub\u003e(obs)\u003c/sub\u003e \u0026ndash; \u0026sum;DIF\u003csub\u003e(sim)\u003c/sub\u003e/nperm)/sd(\u0026sum;DIF\u003csub\u003e(sim)\u003c/sub\u003e), where \u0026sum;DIF\u003csub\u003e(obs)\u003c/sub\u003e is the sum of observed within dyad dissimilarities, \u0026sum;DIF\u003csub\u003e(sim) \u003c/sub\u003eis the sum of simulated within-dyad dissimilarities, nperm is number of permutations, and sd is standard deviation. A simulation study was conducted to test the robustness of this permutation routine to false positives. First, using the R package AHMbook\u003csup\u003e54\u003c/sup\u003e, we simulated 1,000 random communities of 1,000 species mimicking father-partner dyads and thus consisting of two sample groups of 60 samples each. No \u003cem\u003ea prior\u003c/em\u003ei driver of correlation in community composition was specified within the 60 dyads. The frequency of false positive results estimated, based on these simulated datasets and the permutation routine described above, was low at 4.6% (Extended Data Fig. 1e). Analyses on predicted metagenome pathways utilized most of the above-described approaches, with two distinctions. First, instead of abundance matrix rarefaction, functional pathways abundances were normalized by their transformation to proportions in each sample. Second, we used only relative-based (i.e., Bray-Curtis) dissimilarities in all analyses on predicted metagenomes. All statistical calculations and data visualizations were conducted using R 4.0.2\u003csup\u003e55\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eChemical analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSample extraction\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe sampling for analyses of axillary chemical profiles was performed by the participants themselves as described above. The samples were transported and stored at -20\u0026deg;C until extraction, which took place as soon as possible after the reception of samples. Left and right axillary samples were successfully collected from all participants (67 father-partner dyads). To allow multiple analyses and storage of the irreplaceable samples, we selected the solvent extraction of the samples, in spite of the drawbacks inherent to this method when compared with direct thermal desorption, especially the lower representation of highly volatile and low molecular mass compounds. The analytes were extracted from the cotton swabs with 400 \u0026micro;l of distilled \u003cem\u003en\u003c/em\u003e-hexane for 24 hours at 8\u0026deg;C plus an additional 10 minutes of sonication. The extracts were transferred into clean vials, two internal standards (10 ng of 1-bromononane and 100 ng of 1-bromoeicosane) were added, and the extracts were either stored at -80\u0026deg;C or directly analysed.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eInstrument setup\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe samples were analysed using a GC\u0026times;GC-TOFMS instrument Pegasus 4D (LECO Corp., St Joseph, MI, USA). The system is based on a quad-jet modulator with two cold liquid-nitrogen jets and two hot-air jets, which trap and refocus the analytes eluting from the first dimension column. The column set consisted of a nonpolar Rxi-5Sil MS (30 m, ID 0.25 mm, d\u003csub\u003ef\u003c/sub\u003e 0.25 \u0026micro;m, Restek, Bellefonte, PA, USA) in the first dimension and a mid-polar Rxi-17Sil MS (1.5 ms, id 0.1 mm, d\u003csub\u003ef\u003c/sub\u003e 0.1 \u0026micro;m, Restek, Bellefonte, PA, USA) in the second dimension. The columns were connected using SilTite\u0026reg; \u0026micro;-Union (Trajan Scientific, Australia). The modulation period was 5 s and the hot pulse duration 600 ms. The temperature program was 50\u0026deg;C (1 min), 8\u0026deg;C/min ramp to 320\u0026deg;C (20 min). The secondary column was set 10\u0026deg;C higher. The modulator offset was 20\u0026deg;C. Helium was used as a carrier gas at a constant flow rate of 1 ml/min. Total run time was 54.75 min. 1\u0026micro;l of the extract was injected into a split/splitless injector heated to 200\u0026deg;C in splitless mode. Acquisition solvent delay was set at 400 s, while MS transfer line temperature was 280\u0026deg;C. The ion source temperature was 220\u0026deg;C and electron impact energy was -70 eV. Collected mass range was 29-600 amu and acquisition rate was 100 scans/s. The detector voltage was 1,500 V.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMeasurement sequence \u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe samples were analysed in randomized order and injected using a 7683 Series Injector (Agilent Technologies). Before each sample, a blank was run using the same temperature program as for the sample runs. To guarantee the reliability of the chemical analyses and to avoid biases in chromatograms alignment, we only considered the samples which were analysed during a single measurement sequence, uninterrupted by any change in the analytical setup, such as column exchange, etc. The only intervention in the setup during the sequence measurement was the regular exchange of the split/splitless liner (Topaz liner, Splitless, Single Taper Gooseneck w/Wool, 4mm \u0026times; 6.5 \u0026times;78.5, Restek), which was replaced after every 10 samples. The longest successful uninterrupted sequence included both left and right axillary samples of 42 father-partner dyads (168 samples), which were then used for data analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData processing\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor peak detection, alignment and quantification, we processed the data generated by the operating software LECO ChromaTOF software v. 4.72 (LECO, St. Joseph, MI) with the development version of ChromaTOF-Tile (v. 0.27). This software applies the recently developed tile-based algorithm for GC\u0026times;GC-TOFMS multiple chromatogram processing, which provides an effective alternative to the widespread pixel-based and peak table-based methods\u003csup\u003e56,57\u003c/sup\u003e. Tile-based algorithm offers two main advantages over the traditional approaches: it is computationally fast and it significantly reduces the problem of 2D retention time misalignments. ChromaTOF-Tile is primarily designed for supervised comparison of two or multiple pre-defined classes (groups) of samples. Therefore, for the purposes of unsupervised exploration of chemical profiles across all body odour samples with no \u003cem\u003e\u0026agrave; priori\u003c/em\u003e classification, we defined all measured samples as one group of 168 samples. To generate the second, reference (control) group, we used the chromatogram of a hexane solution of n-alkanes (C\u003csub\u003e7\u003c/sub\u003e‒C\u003csub\u003e40\u003c/sub\u003e, 1ng/\u0026micro;l each, Sigma Aldrich), obtained using identical chromatographic conditions as the samples. Upon importing into ChromaTOF-Tile, we multiplicated this chromatogram into 16 files with variable dilution factors (0.96‒1.03) to give rise to the reference group with identical and low coefficients of variation for all analytes.\u003c/p\u003e\n\u003cp\u003eTo reduce the number of multiple thousand putative analytes initially detected by ChromaTOF-Tile, operated over the range of masses m/z 29‒600, to the most informative peaks characterizing body odour samples, we took advantage of F-ratio comparison between the two predefined classes calculated by ChromaTOF-Tile for each analyte\u0026rsquo;s peak volumes. At the same time, we optimized the signal-to-noise ratio (S/N ratio) for peak inclusion to eliminate artifact peaks. Both parameters (i.e., F-ratio and S/N ratio) were fine-tuned during the pilot study with six male participants sampled six times each (Supplementary information). Desirable separation of chemical profiles of the six participants and optimum clustering of the six samples per participant were obtained for F-ratio = 2 and S/N ratio = 50 (Extended data Fig. 5). These inclusion parameters were thus used for the main analysis. The resulting dataset was then manually curated to exclude the remaining column bleeding signals and artifacts, peak tailing-related hits, and analytes identified in the extracts of the holding sticks of the cotton swabs used for sampling.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAnalytes identification\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor tentative identification of individual analytes included in the main analysis, both MS spectra and 2D retention characteristics were used. Electron ionization MS spectrum (EIMS) of each compound was compared with mass spectral libraries NIST 2017, Wiley Registry\u0026trade; of Mass Spectral Data (8\u003csup\u003eth\u003c/sup\u003e edition) and an in-house wax ester library\u003csup\u003e58\u003c/sup\u003e. EIMS spectra of all analytes were manually inspected, eventual erroneous library hits corrected, and new identifications proposed based on diagnostic fragments. In parallel, retention indices (RI) for both GC dimensions were retrieved from ChromaTOF-Tile output data and manually compared with NIST and other sources\u003csup\u003e59,60\u003c/sup\u003e. Tentative identification is provided only for compounds matching by both the EIMS and RI data (Supplementary Table Dataset 2).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe peak volumes for all included analytes across the 168 samples were square root transformed to reduce heteroscedasticity and related to the sum of peak volumes of each sample to correct for the differences in sample intensities. For subsequent calculations, three datasets with 84 samples each were prepared and independently processed. The first two included only right or only left axillary samples, while the third consisted of averaged values from left and right axillary sample for each analyte. Each dataset was subjected to Principal Component Analysis with autoscaling and centering, performed in Canoco for Windows v. 4.52. The factor scores for the first three principal components were plotted and used to calculate the matrix of Euclidean distances among all possible pairs of participants. These distances were then used as indicators of chemical dissimilarity among different samples. The differences between within-dyad father-partner distances and averaged distances of the corresponding father to all extra-pair partners were compared with zero expected under the null hypothesis of no influence of the pairing on chemical similarities, using Wilcoxon Signed Rank tests. Independently, we tested whether the observed within-dyad father-partner distances differ from distances obtained under random pairing of fathers with partners. To do so, we generated datasets of randomly paired father and partners (999 permutations) and compared the observed average father-partner distance with the distribution of average father-partner distances in the permuted datasets.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eReporting summary\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFurther information on research design is available in the Nature Portfolio Reporting Summary linked to this article.\u003c/p\u003e\n\u003cp\u003e31. Sch\u0026auml;fer, L., Schriever, V. A. \u0026amp; Croy, I. Human olfactory dysfunction: causes and consequences. \u003cem\u003eCell Tissue Res.\u003c/em\u003e \u003cstrong\u003e383\u003c/strong\u003e, 569\u0026ndash;579 (2021).\u003c/p\u003e\n\u003cp\u003e32. Kohoutov\u0026aacute;, D., Rube\u0026scaron;ov\u0026aacute;, A. \u0026amp; Havl\u0026iacute;ček, J. Shaving of axillary hair has only a transient effect on perceived body odor pleasantness. \u003cem\u003eBehav. Ecol. Sociobiol.\u003c/em\u003e \u003cstrong\u003e66\u003c/strong\u003e, 569\u0026ndash;581 (2012).\u003c/p\u003e\n\u003cp\u003e33. Havlicek, J. \u0026amp; Lenochova, P. Environmental Effects on Human Body Odour. in \u003cem\u003eChemical Signals in Vertebrates XI\u003c/em\u003e (eds. Hurst, J. L., Beynon, R. J., Roberts, S. C. \u0026amp; D., W. T.) vol. 11 199\u0026ndash;212 (Springer, 2008).\u003c/p\u003e\n\u003cp\u003e34. Fialov\u0026aacute;, J., Roberts, S. C. \u0026amp; Havl\u0026iacute;ček, J. Consumption of garlic positively affects hedonic perception of axillary body odour. \u003cem\u003eAppetite\u003c/em\u003e \u003cstrong\u003e97\u003c/strong\u003e, 8\u0026ndash;15 (2016).\u003c/p\u003e\n\u003cp\u003e35. Havl\u0026iacute;ček, J., Lenochov\u0026aacute;, P., Oberzaucher E., Grammer, K. \u0026amp; Roberts, S. C. Does length of sampling affects quality of body odor samples? \u003cem\u003eChemosens. Percept.\u003c/em\u003e (2011).\u003c/p\u003e\n\u003cp\u003e36. Lenochova, P., Roberts, S. C. \u0026amp; Havlicek, J. Methods of Human Body Odor Sampling: The Effect of Freezing. \u003cem\u003eChem. Scences\u003c/em\u003e \u003cstrong\u003e34\u003c/strong\u003e, 127\u0026ndash;138 (2009).\u003c/p\u003e\n\u003cp\u003e37. Arrindell, W. A. \u003cem\u003eet al.\u003c/em\u003e The development of a short form of the EMBU 1: Its appraisal with students in Greece, Guatemala, Hungary and Italy. \u003cem\u003ePers. Individ. Dif.\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 613\u0026ndash;628 (1999).\u003c/p\u003e\n\u003cp\u003e38. \u0026Scaron;affa, G., \u0026Scaron;těrbov\u0026aacute;, Z. \u0026amp; Prokop, P. Parental Investment Is Biased toward Children Named for Their Fathers. \u003cem\u003eHum. Nat.\u003c/em\u003e \u003cstrong\u003e32\u003c/strong\u003e, 387\u0026ndash;405 (2021).\u003c/p\u003e\n\u003cp\u003e39. Fraley, R. C., Heffernan, M. E., Vicary, A. M. \u0026amp; Brumbaugh, C. C. The Experiences in Close Relationships-Relationship Structures Questionnaire: A Method for Assessing Attachment Orientations Across Relationships. \u003cem\u003ePsychol. Assess.\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 615\u0026ndash;625 (2011).\u003c/p\u003e\n\u003cp\u003e40. B\u0026uuml;rkner, P. C. brms: An R package for Bayesian multilevel models using Stan. \u003cem\u003eJ. Stat. Softw.\u003c/em\u003e \u003cstrong\u003e80\u003c/strong\u003e, 1\u0026ndash;28 (2017).\u003c/p\u003e\n\u003cp\u003e41. Glenn, T. C. \u003cem\u003eet al.\u003c/em\u003e Adapterama I: universal stubs and primers for 384 unique dual-indexed or 147,456 combinatorially-indexed Illumina libraries (iTru \u0026amp;amp; iNext). \u003cem\u003ePeerJ\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, e7755 (2019).\u003c/p\u003e\n\u003cp\u003e42. Klindworth, A. \u003cem\u003eet al.\u003c/em\u003e Evaluation of general 16S ribosomal RNA gene PCR primers for classical and next-generation sequencing-based diversity studies. \u003cem\u003eNucleic Acids Res.\u003c/em\u003e \u003cstrong\u003e41\u003c/strong\u003e, 1\u0026ndash;11 (2013).\u003c/p\u003e\n\u003cp\u003e43. Jiang, H., Lei, R., Ding, S. W. \u0026amp; Zhu, S. Skewer: A fast and accurate adapter trimmer for next-generation sequencing paired-end reads. \u003cem\u003eBMC Bioinformatics\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 1\u0026ndash;12 (2014).\u003c/p\u003e\n\u003cp\u003e44. Callahan, B. J. \u003cem\u003eet al.\u003c/em\u003e DADA2: High-resolution sample inference from Illumina amplicon data. \u003cem\u003eNat. Methods\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 581\u0026ndash;583 (2016).\u003c/p\u003e\n\u003cp\u003e45. Edgar, R. C., Haas, B. J., Clemente, J. C., Quince, C. \u0026amp; Knight, R. UCHIME improves sensitivity and speed of chimera detection. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 2194\u0026ndash;2200 (2011).\u003c/p\u003e\n\u003cp\u003e46. Wang, Q., Garrity, G. M., Tiedje, J. M. \u0026amp; Cole, J. R. Na\u0026iuml;ve Bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy. \u003cem\u003eAppl. Environ. Microbiol.\u003c/em\u003e \u003cstrong\u003e73\u003c/strong\u003e, 5261\u0026ndash;5267 (2007).\u003c/p\u003e\n\u003cp\u003e47. Quast, C. \u003cem\u003eet al.\u003c/em\u003e The SILVA ribosomal RNA gene database project: Improved data processing and web-based tools. \u003cem\u003eNucleic Acids Res.\u003c/em\u003e \u003cstrong\u003e41\u003c/strong\u003e, 590\u0026ndash;596 (2013).\u003c/p\u003e\n\u003cp\u003e48. Pafčo, B. \u003cem\u003eet al.\u003c/em\u003e Metabarcoding analysis of strongylid nematode diversity in two sympatric primate species. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 1\u0026ndash;11 (2018).\u003c/p\u003e\n\u003cp\u003e49. Douglas, Gavin, M. \u003cem\u003eet al.\u003c/em\u003e PICRUSt2: An improved and extensible approach for metagenome inference. \u003cem\u003ebioRxiv\u003c/em\u003e 672295 (2020).\u003c/p\u003e\n\u003cp\u003e50. Ye, Y. \u0026amp; Doak, T. G. A parsimony approach to biological pathway reconstruction/inference for genomes and metagenomes. \u003cem\u003ePLoS Comput. Biol.\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 1\u0026ndash;8 (2009).\u003c/p\u003e\n\u003cp\u003e51. Brooks, M. E. \u003cem\u003eet al.\u003c/em\u003e glmmTMB balances speed and flexibility among packages for zero-inflated generalized linear mixed modeling. \u003cem\u003eR J.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 378\u0026ndash;400 (2017).\u003c/p\u003e\n\u003cp\u003e52. Storey, J. D. \u0026amp; Tibshirani, R. Statistical significance for genomewide studies. \u003cem\u003eProc. Natl. Acad. Sci. U. S. A.\u003c/em\u003e \u003cstrong\u003e100\u003c/strong\u003e, 9440\u0026ndash;9445 (2003).\u003c/p\u003e\n\u003cp\u003e53. Gotelli, N. J. \u0026amp; McCabe, D. J. Species co-occurrence: A meta-analysis of J. M. Diamond\u0026rsquo;s assembly rules model. \u003cem\u003eEcology\u003c/em\u003e \u003cstrong\u003e83\u003c/strong\u003e, 2091\u0026ndash;2096 (2002).\u003c/p\u003e\n\u003cp\u003e54. Kery, M. \u0026amp; Royle, J. A. \u003cem\u003eApplied Hierarchical Modeling in Ecology: Analysis of distribution, abundance and species richness in R and BUGS: Volume 2: Dynamic and Advanced Models.\u003c/em\u003e (Academic Press, 2020).\u003c/p\u003e\n\u003cp\u003e55. R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing. (2020).\u003c/p\u003e\n\u003cp\u003e56. Marney, L. C. \u003cem\u003eet al.\u003c/em\u003e Tile-based Fisher-ratio software for improved feature selection analysis of comprehensive two-dimensional gas chromatography-time-of-flight mass spectrometry data. \u003cem\u003eTalanta\u003c/em\u003e \u003cstrong\u003e115\u003c/strong\u003e, 887\u0026ndash;895 (2013).\u003c/p\u003e\n\u003cp\u003e57. Parsons, B. A. \u003cem\u003eet al.\u003c/em\u003e Tile-Based Fisher Ratio Analysis of Comprehensive Two-Dimensional Gas Chromatography Time-of-Flight Mass Spectrometry (GC \u0026times; GC-TOFMS) Data Using a Null Distribution Approach. \u003cem\u003eAnal. Chem.\u003c/em\u003e \u003cstrong\u003e87\u003c/strong\u003e, 3812\u0026ndash;3819 (2015).\u003c/p\u003e\n\u003cp\u003e58. Urbanov\u0026aacute;, K., Vrkoslav, V., Valterov\u0026aacute;, I., H\u0026aacute;kov\u0026aacute;, M. \u0026amp; Cvačka, J. Structural characterization of wax esters by electron ionization mass spectrometry. \u003cem\u003eJ. Lipid Res.\u003c/em\u003e \u003cstrong\u003e53\u003c/strong\u003e, 204\u0026ndash;213 (2012).\u003c/p\u003e\n\u003cp\u003e59. Adams, R. P. \u003cem\u003eIdentification of Essential Oil Components by Gas Chromatography/mass Spectrometry\u003c/em\u003e. (Allured Publishing Corporation, 2007).\u003c/p\u003e\n\u003cp\u003e60. Str\u0026aacute;nsk\u0026yacute;, K., Zarev\u0026uacute;cka, M., Valterov\u0026aacute;, I. \u0026amp; Wimmer, Z. Gas chromatographic retention data of wax esters. \u003cem\u003eJ. Chromatogr. A\u003c/em\u003e \u003cstrong\u003e1128\u003c/strong\u003e, 208\u0026ndash;219 (2006).\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-2531352/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-2531352/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Women tend to choose partners who resemble their father in certain characteristics. In non-human mammals, similar parental imprinting-like effects are often odour-mediated1,2, which requires not only learning of parental odour but also similarity between parents and prospective mates in the microbial communities responsible for production of the critical volatile compounds. Here, in light of growing recognition of human capacity for social olfaction3,4, we tested the possibility that women’s preferences are shaped by paternal phenotype. Axillary body odour samples from women’s fathers and romantic male partners were evaluated for similarity at the perceptual level as well as in both their microbiome diversity (using 16S rDNA sequencing) and chemical composition (using GC×GC-TOFMS). We found that partners’ body odour was perceived as more similar to paternal odour than would be expected by chance. In parallel, similarities in both microbiota composition and chemical profiles were significantly higher in partner-father dyads than in randomly generated dyads. While women’s relationship quality with the father did not predict father-partner similarity in microbial or chemical profiles, it was positively associated with perceived odour similarity. Our results thus reveal a suite of mechanisms by which parental odour cues can serve as a template in human mate choice.","manuscriptTitle":"Women choose romantic partners resembling their father in body odour","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2023-01-31 19:34:26","doi":"10.21203/rs.3.rs-2531352/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"7ba36559-e827-49d8-8e3e-a7f29e4fe1c0","owner":[],"postedDate":"January 31st, 2023","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":18843044,"name":"Biological sciences/Psychology/Human behaviour"},{"id":18843045,"name":"Biological sciences/Microbiology/Microbial communities/Microbiome"},{"id":18843046,"name":"Biological sciences/Chemical biology/Chemical ecology"}],"tags":[],"updatedAt":"2023-02-17T08:55:46+00:00","versionOfRecord":[],"versionCreatedAt":"2023-01-31 19:34:26","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-2531352","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-2531352","identity":"rs-2531352","version":["v1"]},"buildId":"-HB7Z8yhvgn0wM9Nzuekk","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}