eXclusionarY: Ten years later, where are the sex chromosomes in GWAS?

Abstract

Ten years ago, a detailed analysis of genome-wide association studies showed that only 33% of the studies included the X chromosome. Multiple recommendations were made to combat eXclusion. Here we re-surveyed the research landscape to determine if these earlier recommendations had been translated. Unfortunately, among the summary statistics reported in 2021 in the NHGRI-EBI GWAS catalog, only 25% provided results for the X chromosome and 3% for the Y chromosome, suggesting that the eXclusion phenomenon documented earlier not only persists but has also expanded into an eXclusionarY problem. Normalizing by physical length of the chromosome, the average number of studies published until 11/29/22 with genome-wide significant findings on the X chromosome is ~1 study/Mb. In contrast, it ranges from ~6 to ~16 studies/Mb for chromosomes 4 and 19, respectively. Compared with the autosomal growth rate of ~0.086 studies/Mb/year over the last decade, studies of the X chromosome grew at less than one-seventh that rate, only ~0.012 studies/Mb/year. Among the studies that reported significant association on the X chromosome, there were extreme heterogeneities in how they analyzed the data and documented the results, suggesting the need for guidelines. Not surprisingly, among the 430 scores sampled from the PolyGenic Score catalog, 0% contained weights for sex chromosomal SNPs. To overcome the dearth of sex chromosome analyses, we provide five sets of recommendations and future directions. Finally, until the sex chromosomes are included in a whole-genome study, instead of GWAS, we propose they be more properly referred to as "AWAS" for "autosome-wide scan".

Concerning the eXclusion in human genomics: The choice of sex chromosome representation in the human genome drastically affects number of identified variants

“…The introduction of read mapping errors on the X chromosome only affects 5% of the total length of the X chromosome, which equates to only 0.25% of the variants called become unreliable when not accounting for sex chromosome complement and using a Default reference genome (Figure 1; Table 2). Thus, the common practice of purposefully introducing an error rate of 5% (excluding the X chromosome) to potentially avoid an error rate of 0.25% (including the X chromosome) is excessive and, technically speaking, precludes the use of the term “genome-wide” in most association studies in humans (Sun et al 2023; Wise et al 2013). However, it is a relatively trivial task to inform the reference genome with the sex chromosome complement when mapping samples and accommodate changes in ploidy across different regions; thus ensuring that reliable variant calls across, even within the PARs and XTR (Carey et al 2022; Webster et al 2019).…”

“…In line with prior work, we provided additional evidence that technical artifacts of including the sex chromosomes in genomics analyses can be negated with available information and tools (Olney et al 2019; Webster et al 2019). We know that, though the ‘eXclusion’ of the X chromosome is widespread (Wise et al 2013), the exclusion of the Y is even more extensive in empirical and clinical genomics (Sun et al 2023). SCC-aware reference genomes can effectively negate the effects of homology on the sex chromosomes in XX individuals and reduce this mis-mapping in XY individuals, allowing for their accurate inclusion in human genomics studies (Oill, 2022).…”

“…Although this is well understood biologically, it introduces technical artifacts within modern genomic analyses that require correction to prevent potentially erroneous conclusions (Carey et al 2022; Webster et al 2019). Though these technical artifacts have remained ignored in many empirical and clinical studies, they have been used as justification to ignore the sex chromosomes on a grand scale and, therefore, the importance of sex-linked variation to human health is likely greatly underestimated (Inkster et al 2023; Khramtsova et al 2019; Köferle et al 2022; Natri et al 2019; Sun et al 2023; Wise et al 2013). Here, we aim to better grasp the scope of data lost by excluding or misrepresenting the sex chromosomes in human genomics.…”

“…Here, we aim to better grasp the scope of data lost by excluding or misrepresenting the sex chromosomes in human genomics. We urge empiricists and clinicians to confront these issues moving forward to simultaneously increase the number of genome-wide association studies (GWAS) and reduce the numbers of autosome-wide association scan studies (AWAS) currently being published (Sun et al 2023).…”

“…Thus, many published studies in humans blatantly ignore 5% or more of the human genome when conducting routine genomic analyses (Koboldt, 2020; Wise et al 2013; Zverinova and Guryev, 2021). Indeed, despite recent advances in methodology to control for known technical artifacts inherent when analyzing the sex chromosomes (e.g., Webster et al 2019), little progress has been made to further incorporate the X chromosome into broader biological analyses (Carey et al 2022; Sun et al 2023; Wise et al 2013).…”