Dosage Compensation Is Assumed, Not Modelled, in Sex-Stratified Lipid GWAS

Question: How do sex-stratified and sex-interaction GWAS of lipid traits in biobanks such as UK Biobank and MVP account for X-chromosome dosage compensation and cross-sex genetic correlation, and what loci have they replicated beyond APOE?

Answer

Across the retrieved literature, X-chromosome dosage compensation enters sex-stratified lipid GWAS almost entirely as a fixed modelling assumption rather than an estimated quantity, while cross-sex genetic correlation is reported as a genome-wide scalar that is now known to mask locus-level dimorphism. The assumption most often adopted — near-complete X inactivation, with males coded hemizygous and females diploid — receives its strongest empirical support from chrX heritability analyses showing a pronounced male bias in chrX heritability consistent with near-complete dosage compensation, alongside larger active allele effects on chrX than autosomes [20]. Toolkits now formalise this by offering sex-stratified and sex-combined XWAS with XCI-aware coding and tests of sex differences in heritability and genetic correlation [5]. But the escape-from-XCI literature shows that female-biased X expression is a poor proxy for escape and that escape varies by gene and tissue, so a single genome-wide dosage model cannot be correct at every locus [6], and inactive-X expression is itself a strong linear predictor of female–male expression differences and of apparent additive versus dominance effects [11]. On the replication question, the corpus does not support a confident answer: the sex-heterogeneous lipid loci nominated in biobank-scale work beyond APOE/APOC1 are few, autosomal, and modest in effect — CELSR2, GPAM, PLTP, ZPR1 and CMIP [1] — with exactly one X-linked LDL-C signal reported from a male-stratified XWAS, at PTCHD1-AS [17]. Taken together, these papers describe a field that has solved the plumbing of sex-aware association testing without yet modelling dosage compensation or cross-sex architecture at the resolution lipids would require.

X-chromosome dosage compensation is assumed rather than estimated, and the assumption is calibrated only at the level of total heritability

The strongest quantitative anchor for the standard assumption comes from 48 quantitative traits in 343,695 UK Biobank participants with replication in 412,181 FinnGen participants, where chrX contributes heritability at roughly 3% of the autosomal rate, and where a pronounced male bias in chrX heritability supports near-complete dosage compensation between the sexes through X inactivation [20]. The same analysis reports subtle evidence that escape from XCI contributes to human height, and shows that under an assumption of full XCI the chrX contribution exceeds expectation, attributable to systematically larger active allele effects on chrX than on autosomes in both sexes [20]. This is a population-average calibration, not a per-locus dosage model, and the mechanistic literature argues that a single model is inadequate: XCI status has now been directly determined for 380 X-linked genes across 30 tissues, including 198 determined for the first time, providing the reference needed to dissect sex-biased traits [9]. Escape magnitude, not female-biased expression, predicts sex-biased X-linked expression, and XCI extends into the pseudoautosomal region so that the extent of inactivation predicts male-biased expression of PAR genes [6]. In females, inactive-X expression is a strong linear predictor of female–male expression differences, and variants that influence XCI skew can generate apparent dominance effects on traits including BMI, estradiol and testosterone [11]. Practical implementations have followed: GXwasR performs sex-combined and sex-stratified GWAS and XWAS while accounting for XCI, and adds tests for sex differences in SNP effects and heritability plus genetic correlation estimation [5], and geneXplore exposes XWAS summary statistics across 1,944 phenotypes while separating random XCI, escape from XCI, and sex-stratified analyses [21].

Genome-wide cross-sex genetic correlation is a weak summary of lipid genetic architecture

Average metrics of cross-sex genetic correlation and heritability can mask substantial variability between loci: using local genetic correlation analysis across 157 UK Biobank quantitative traits, nearly every trait shows evidence for sex-dimorphic effects in at least one locus, and only the standardised scale of genetic effects is informative for heritability comparisons [7]. Decomposition of sex-different phenotypic correlations into genetic and environmental components found 250 trait pairs with significantly sex-different phenotypic correlations, none attributable solely to sex-specific genetic correlation, with triglycerides among the traits proposed to reflect low between-sex genetic correlation [3]. In a multivariate analysis of 17 metabolic traits, the genome-wide between-sex genetic covariance matrix showed only 1.15% asymmetry, with sexually concordant variation exceeding sexually antagonistic variation by 18.5-fold, while local regions were far more heterogeneous [18]. At the level of individual loci, comparing allelic effect directions between sex-stratified analyses in UK Biobank and the Hybrid Mouse Diversity Panel showed high cross-sex concordance even for alleles not reaching nominal significance, consistent with thousands of loci regulating each human trait [4]. The practical implication for lipids is that a single reported cross-sex genetic correlation near unity is compatible with genuine locus-level dimorphism.

For LDL-C, heritability shows no sex difference, which relocates sex interaction to variance components genetic models do not fit

The clearest direct test in the retrieved corpus is a sex-stratified GWAS, XWAS, GxSex and heritability analysis of LDL-C and statin treatment in All of Us, which reports that heritability and additive genetic variability of LDL-C show no sex differences while residual and phenotypic variability do, suggesting environmental factors drive variation in LDL-C and statin treatment rates across sex and race groups [17]. That study also documents that women are treated with statins less often and appear to respond less effectively, with the highest statin-treated median LDL-C among Black women [17]. Treatment asymmetry is a direct threat to sex-interaction lipid GWAS: sex-specific pQTL Mendelian randomisation against lipid summary statistics identified 82 female-exclusive and 82 male-exclusive causal protein–lipid links plus 38 with significantly different magnitudes, and the authors replicated 16 sex-specific and 4 sex-differentiated relationships in Lifelines including GWAS restricted to non-statin users to rule out confounding by lipid-lowering medication more often prescribed in males [2].

Beyond APOE/APOC1, the replicated sex-heterogeneous lipid loci form a short autosomal list plus one X-linked signal

Sex-stratified GWAS of four lipid profiles in 124,920 All of Us whole-genome-sequenced participants of diverse ancestry recovered five genes with significant sex-heterogeneous effects: CELSR2 with a stronger HDL-C association in males (β female −0.022 versus β male −0.045), and GPAM (β female 0.042 versus β male 0.013), PLTP (0.06 versus 0.02) and ZPR1 (0.050 versus 0.014) stronger in females, with CMIP stronger in females for triglycerides (0.037 versus 0.019) [1]. These are reported as sex-heterogeneous effects in a single cohort rather than as independent replications of prior sex-specific loci. On the X chromosome, sex-stratified GWAS in All of Us recapitulated known sex-differentiated effects at APOE/APOC1, and a male-stratified XWAS identified an LDL-C-associated SNP in the PTCHD1-AS region [17] — to my reading the only X-linked lipid locus surfaced in this corpus. Related sex-differential lipid genetics outside the lipid-GWAS frame includes a glucocorticoid receptor coding variant, rs6190, associated with increased cholesterol in women in both UK Biobank and All of Us and shown in mice to transactivate Pcsk9 and Bhlhe40 [16].

Many reported sex-interaction lipid signals are plausible scale or technical artefacts rather than biology

Phenotype scale alone explains a large share of apparent sex interaction: across 32 UK Biobank phenotypes, a power-transformation scaling explained 46% of PGS×Sex interactions and a simple log transformation explained 23%, with GxSex effects on height vanishing on the log scale while log scaling uncovered biologically meaningful female-specific testosterone effects [13]. Sex-specific signals can also be mapping artefacts: in a meta-analysis of 1,922 case-parent trios, four genome-wide-significant sex-specific signals that passed standard QC proved to be sequencing errors from mis-mapped reads with sequence similarity to the sex chromosomes [15]. Sex-chromosome genotype data require dedicated handling — genotype-level QC in UK Biobank and All of Us demonstrated the need for sex-specific X-chromosome filtering [8], sex-complement-informed alignment and accurate ploidy parameters increase true positives in the PARs and reduce false positives in XY samples [14], and Hardy-Weinberg testing on the X can show inflated type I error when sex differences in minor allele frequency are present [12].

Conflicts and gaps

The central gap is bibliographic: no sex-stratified or sex-interaction lipid GWAS in the Million Veteran Program appears in the retrieved corpus. MVP contributes to this literature through a sex-chromosome trisomy PheWAS across 1.5 million participants in MVP, FinnGen and UK Biobank, where 47,XXY, 47,XYY and 47,XXX showed overlapping phenotypes including vascular disease, suggesting shared X/Y homolog gene effects [10], and through a GWAS of antipsychotic-induced change in total cholesterol in 59,372 MVP participants that is not sex-stratified [19]. Claims about MVP-specific lipid handling therefore cannot be evaluated from these papers.

A second conflict concerns the appropriate null for dosage compensation. Fu and colleagues' conclusion of near-complete compensation rests on a male-biased chrX heritability pattern [20], while work on inactive-X expression and escape argues that escape magnitude varies enough across genes and tissues to generate measurable sex-biased expression and dominance-like effects [6] [11]. Both can be true at different genomic scales, but no retrieved paper tests a locus-specific dosage model against lipid phenotypes.

A third tension is between evidence that sex-dimorphic loci are near-ubiquitous locally [7] and reports that LDL-C heritability and additive genetic variance show no sex difference [17] or that between-sex covariance is almost symmetric [18]. Whether this reflects genuine biology, scale dependence [13], or power differences is unresolved.

What would change the answer

A sex-stratified lipid GWAS in MVP with chrX included, using XCI-aware coding and reporting local rather than only global cross-sex genetic correlation, would directly fill the missing cell. Feeding the directly determined XCI status of 380 genes across 30 tissues [9] into an XWAS as a per-gene prior — instead of a single global inactivation assumption — and benchmarking it against the toolkit implementations now available [5] [21] would convert dosage compensation from an assumption into a tested parameter. Replication of the CELSR2, GPAM, PLTP, ZPR1 and CMIP sex-heterogeneous effects [1] in an independent biobank with sex-stratified effect estimates and a pre-registered phenotype scale would establish whether they survive the scaling critique [13]. Finally, statin-stratified or treatment-naive analyses, as already implemented in the sex-specific pQTL work [2], together with sex-specific X-chromosome genotyping QC and sex-complement-informed variant calling [8] [14], would separate sex-differential lipid biology from sex-differential treatment and technical noise.

References

[1] Li, Y., Lee, I.-H., Kong, S. W.. "A sex-specific genome-wide association study of blood lipid levels in All of Us". medRxiv. 2026-02-27T00:00:00.000Z. URL: https://doi.org/10.1101/2025.11.21.25340771

[2] Zanetti, D., Koprulu, M., Grosso, F., Zhernakova, D. V., Crobu, F., Sanna, S.. "Exploring sex-specific causal links between thousands of proteins and lipid metabolism using the UK Biobank Pharma proteomics data.". medRxiv. 2025-09-18T00:00:00.000Z. URL: https://doi.org/10.1101/2025.09.16.25335948

[3] Fritz, A., Darrous, L., Bonnelykke, K., Pedersen, A. G., Kutalik, Z.. "Sex-different phenotypic correlations: Due to genes or environment?". medRxiv. 2026-07-15T00:00:00.000Z. URL: https://doi.org/10.64898/2026.07.13.26357694

[4] Miller, A. K., Pan, C., Bartlett, J., Lusis, A. J., Crawford, D. C., Williams, S. M., Buchner, D. A.. "Concordance between male- and female-specific GWAS results helps define underlying genetic architecture of complex traits". bioRxiv. 2024-12-21T00:00:00.000Z. URL: https://doi.org/10.1101/2022.09.28.509932

[5] Bose, B., Bolstein, F., Kim, J., Winters, J., Actkins, K., Mayer, D., Niarchou, M., Congivaram, H. T. S., Edwards, D. R. V., Davis, L., Stranger, B.. "GXwasR: A Toolkit for Investigating Sex-Differentiated Genetic Effects on Complex Traits". medRxiv. 2025-06-11T00:00:00.000Z. URL: https://doi.org/10.1101/2025.06.10.25329327

[6] DeCasien, A. R.. "Escape from X inactivation varies across genes and tissues and shapes sex-biased sex chromosome gene expression". bioRxiv. 2025-12-22T00:00:00.000Z. URL: https://doi.org/10.64898/2025.12.19.695225

[7] Uffelmann, E., de Leeuw, C., Schipper, M., Posthuma, D.. "Local Genetic Sex Differences in Quantitative Traits". bioRxiv. 2025-04-18T00:00:00.000Z. URL: https://doi.org/10.1101/2023.05.04.539410

[8] Chundru, V. K., Wright, H. I. W., Frayling, T. M., Wood, A. R., Weedon, M. N., Wright, C. F., Beaumont, R. N., Hawkes, G.. "Genotype-level quality control substantially reduces error rates in population-scale whole-genome sequencing". bioRxiv. 2025-11-07T00:00:00.000Z. URL: https://doi.org/10.1101/2025.11.06.686913

[9] Gylemo, B., Bensberg, M., Nestor, C. E.. "A whole-organism landscape of X-inactivation in humans". bioRxiv. 2025-03-26T00:00:00.000Z. URL: https://doi.org/10.1101/2023.06.26.546519

[10] Davis, S., Liu, A., Teerlink, C. C., Lapato, D. M., Gorman, B. R., Genovese, G., Singh, M., Reeve, M. P., Elswick Gentry, A., Donner, K. M., Sipila, T. P., Ghazal, A., Pagadala, M., Panizzon, M. S., Lancaster, E. E., FinnGen,, UKB working group,, Chatzinakos, C., Ganna, A., Bigdeli, T., Daly, M. J., Lynch, J., Ross, J. L., Peterson, R., Hauger, R. L.. "Prevalence and disease risks for male and female sex chromosome trisomies: a registry-based phenome-wide association study in 1.5 million participants of MVP, FinnGen, and UK Biobank". medRxiv. 2025-02-02T00:00:00.000Z. URL: https://doi.org/10.1101/2025.01.31.25321488

[11] Zhu, C., Xu, L., Harpak, A.. "Escape from X inactivation drives sex differences in gene expression". bioRxiv. 2025-12-11T00:00:00.000Z. URL: https://doi.org/10.1101/2025.08.01.668097

[12] Zhang, L., Paterson, A. D., Sun, L.. "A General Statistical Framework for Hardy-Weinberg Equilibrium Inference on the X Chromosome". bioRxiv. 2026-05-20T00:00:00.000Z. URL: https://doi.org/10.64898/2026.05.17.725730

[13] Costantino, M., Fonseca, R., Liu, Z., Huang, Z., Sankararaman, S., Mathieson, I., Dahl, A.. "Choice of phenotype scale is critical in biobank-based GxE tests". bioRxiv. 2026-01-21T00:00:00.000Z. URL: https://doi.org/10.64898/2026.01.20.694695

[14] Oill, A. M. T., Plaisier, S. B., Phung, T. N., Wilson, M. A.. "Best practices for improving alignment and variant calling on human sex chromosomes". bioRxiv. 2025-04-30T00:00:00.000Z. URL: https://doi.org/10.1101/2025.04.29.651297

[15] Kanchan, K., ERDOGAN-YILDIRIM, Z., Berke, S. R., Mukhopadhyay, N., Ray, D., Simpson, C. L., Bidinger, J. A., Curtis, S. W., Butali, A., Schwender, H., Scott, A. F., Bailey Wilson, J., Beaty, T. H., Leslie, E., Marazita, M. L., Ruczinski, I.. "A study of sex-specific genetic effects underlying risk of orofacial clefts also highlights the potential impact of sequencing errors due to short read mis-mapping". medRxiv. 2026-07-09T00:00:00.000Z. URL: https://doi.org/10.64898/2026.07.07.26357463

[16] Durumutla, H. B., Haller, A., Noble, G., Prabakaran, A. D., McFarland, K., Latimer, H., Rajput, A., Akinborewa, O., Khales, B. N., Hui, D. Y., Quattrocelli, M.. "The human glucocorticoid receptor variant rs6190 promotes blood cholesterol and atherosclerosis". bioRxiv. 2025-04-30T00:00:00.000Z. URL: https://doi.org/10.1101/2024.11.27.625727

[17] Liquori, J. L., Tian, X., Musharoff, S. A.. "Phenotypic and genetic sex differences in LDL-C and statin treatment in All of Us". medRxiv. 2026-10-19T00:00:00.000Z. URL: https://doi.org/10.64898/2026.01.26.26344867

[18] Chakraborty, S., Mitra, K., Chakrabarty, A.. "In search of intralocus sexual conflict in the multivariate and local genetic architecture of metabolic traits in humans". bioRxiv. 2026-07-01T00:00:00.000Z. URL: https://doi.org/10.64898/2026.06.28.735027

[19] Kazemi, H., Drake, J. E., Bacanu, S.-A., Mcmahon, B. H., Agarwal, K., Zuniga, A. M., Choudhury, S., Abbaszadegan, H., Dhaubhadel, S., Johnson, K. A., Kreyenbuhl, J. A., Marder, S. R., Harvey, P. D., Vladimirov, V. I., Million Veteran Program,, MVP Program Office,, MVP Core Operations,, MVP Steering Committee,, Fanous, A. H.. "Genomic and transcriptomic insights into antipsychotic-induced changes in total cholesterol and body mass index in a multi-ancestry cohort of the US veterans". medRxiv. 2026-07-22T00:00:00.000Z. URL: https://doi.org/10.64898/2026.07.20.26358275

[20] Fu, Y., Kenttamies, A., Ruotsalainen, S. E., Pirinen, M., Tukiainen, T.. "Role of X chromosome and dosage compensation mechanisms in complex trait genetics". bioRxiv. 2025-01-17T00:00:00.000Z. URL: https://doi.org/10.1101/2025.01.16.633321

[21] Cook, N., Boulais-Richard, J., Zeng, Y., Yang, C., Budde, J., Taliun, D., Gagliano Taliun, S. A., Cruchaga, C., Belloy, M. E.. "geneXplore: An Interactive Browser for X Chromosome-Wide Association Study Results". medRxiv. 2026-07-14T00:00:00.000Z. URL: https://doi.org/10.64898/2026.07.14.26357489

How this survey was made

This survey answers one question: “How do sex-stratified and sex-interaction GWAS of lipid traits in biobanks such as UK Biobank and MVP account for X-chromosome dosage compensation and cross-sex genetic correlation, and what loci have they replicated beyond APOE?”

We keep our own collection of recent preprints and refresh it every day from bioRxiv, medRxiv, arXiv, PsyArXiv, SocArXiv and ChemRxiv. For each survey, we search that collection for recent work and keep the strongest matches: papers that use the same terms as the question, and papers on the same topic in different words.

The text was written from the titles and abstracts of those papers; the full texts were not read. Statements that draw on a particular paper are marked with a number in brackets. The reference list just above is the 21 papers whose work this text draws on, chosen from the 97 papers those searches returned; the rest of the shortlist is not printed. It is not a complete review of everything published on the question.

Treat this as a starting point and a reading list, not a peer-reviewed review. The searches behind this survey ran on October 21, 2026 and covered work from the previous two years.