Dataset Information

Genetic analyses identify widespread sex-differential participation bias.

ABSTRACT: Genetic association results are often interpreted with the assumption that study participation does not affect downstream analyses. Understanding the genetic basis of participation bias is challenging since it requires the genotypes of unseen individuals. Here we demonstrate that it is possible to estimate comparative biases by performing a genome-wide association study contrasting one subgroup versus another. For example, we showed that sex exhibits artifactual autosomal heritability in the presence of sex-differential participation bias. By performing a genome-wide association study of sex in approximately 3.3 million males and females, we identified over 158 autosomal loci spuriously associated with sex and highlighted complex traits underpinning differences in study participation between the sexes. For example, the body mass index-increasing allele at FTO was observed at higher frequency in males compared to females (odds ratio = 1.02, P = 4.4 × 10^-³⁶). Finally, we demonstrated how these biases can potentially lead to incorrect inferences in downstream analyses and propose a conceptual framework for addressing such biases. Our findings highlight a new challenge that genetic studies may face as sample sizes continue to grow.

SUBMITTER: Pirastu N

PROVIDER: S-EPMC7611642 | biostudies-literature | 2021 May

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Genetic analyses identify widespread sex-differential participation bias.

Pirastu Nicola N Cordioli Mattia M Nandakumar Priyanka P Mignogna Gianmarco G Abdellaoui Abdel A Hollis Benjamin B Kanai Masahiro M Rajagopal Veera M VM Parolo Pietro Della Briotta PDB Baya Nikolas N Carey Caitlin E CE Karjalainen Juha J Als Thomas D TD Van der Zee Matthijs D MD Day Felix R FR Ong Ken K KK Morisaki Takayuki T de Geus Eco E Bellocco Rino R Okada Yukinori Y Børglum Anders D AD Joshi Peter P Auton Adam A Hinds David D Neale Benjamin M BM Walters Raymond K RK Nivard Michel G MG Perry John R B JRB Ganna Andrea A

Nature genetics 20210422 5

Genetic association results are often interpreted with the assumption that study participation does not affect downstream analyses. Understanding the genetic basis of participation bias is challenging since it requires the genotypes of unseen individuals. Here we demonstrate that it is possible to estimate comparative biases by performing a genome-wide association study contrasting one subgroup versus another. For example, we showed that sex exhibits artifactual autosomal heritability in the pre ...[more]

PMID: 33888908

Similar Datasets

Project description:71 liver RNA samples from three mouse strains - DBA/2J, C57BL/6J and C3H/HeJ - were profiled using a custom-designed microarray monitoring exon and exon-junction expression of 1,020 genes representing 9,406 exons. Gene expression was calculated via two different methods, using the 3'-most exon probe ("3' gene expression profiling") and using all probes associated with the gene ("whole-transcript gene expression profiling"), while exon expression was determined using exon probes and flanking junction probes that spanned across the neighboring exons("exon expression profiling"). Widespread strain and sex influences were detected using a two-way Analysis of Variance (ANOVA) regardless of the profiling method used. However, over 90% of the genes identified in 3' gene expression profiling or whole transcript profiling were identified in exon profiling, along with 75% and 38% more genes, respectively, showing evidence of differential isoform expression. Overall, 55% and 32% of genes, respectively, exhibited strain- and sex-bias differential gene or exon expression. Keywords: Grouped Hybridization material was generated through a random-priming amplification of poly[A]+ purified RNA using primers with a random sequence at the 3' end and a fixed motif at the 5' end that was optimized to generate strand-specific cDNA copies of full-length mRNA transcripts [32]. Since the region used for exon and junction probe selection is constrained to a smaller region, more probes contain sequence with suboptimal characteristics (e.g. high GC content or higher homology to other genes). The hybridization of cDNA, rather than cRNA as commonly done, partially mitigates this issue due to higher specificity and lower background levels [24]. Hybridization conditions were as previously described [33]. All 71 samples were hybridized in a two-channel experiment, where one channel was a common reference, generated by pooling all 71 samples in equal mass. Array hybridizations were done in duplicate with fluor reversal to systematically correct for Cy3/Cy5 dye bias. Array images were processed as described to obtain background noise, single channel intensity and associated measurement error estimates [34]. Expression changes between samples and pool were quantified as mean log10(expression ratio), and associated error.

Project description:ObjectivesLow response rates in epidemiologic research potentially lead to the recruitment of a non-representative sample of controls in case-control studies. Problems in the unbiased estimation of odds ratios arise when characteristics causing the probability of participation are associated with exposure and outcome. This is a specific setting of selection bias and a realistic hazard in many case-control studies. This paper formally describes the problem and shows its potential extent, reviews existing approaches for bias adjustment applicable under certain conditions, compares and applies them.MethodsWe focus on two scenarios: a characteristic C causing differential participation of controls is linked to the outcome through its association with risk factor E (scenario I), and C is additionally a genuine risk factor itself (scenario II). We further assume external data sources are available which provide an unbiased estimate of C in the underlying population. Given these scenarios, we (i) review available approaches and their performance in the setting of bias due to differential participation; (ii) describe two existing approaches to correct for the bias in both scenarios in more detail; (iii) present the magnitude of the resulting bias by simulation if the selection of a non-representative sample is ignored; and (iv) demonstrate the approaches' application via data from a case-control study on stroke.FindingsThe bias of the effect measure for variable E in scenario I and C in scenario II can be large and should therefore be adjusted for in any analysis. It is positively associated with the difference in response rates between groups of the characteristic causing differential participation, and inversely associated with the total response rate in the controls. Adjustment in a standard logistic regression framework is possible in both scenarios if the population distribution of the characteristic causing differential participation is known or can be approximated well.

Dataset Information

Genetic analyses identify widespread sex-differential participation bias.

Publications

Genetic analyses identify widespread sex-differential participation bias.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets