Slope-Hunter: A robust method for index-event bias correction in genome-wide association studies of subsequent traits

Mahmoud, Osama; Dudbridge, Frank; Smith, George Davey; Munafò, Marcus R.; Tilling, Kate

doi:10.1101/2020.01.31.928077

Cited by 21 publications

(25 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Inverse-probability-of-sampling-weighted (IPW) regression has been applied to achieve unbiased estimates from analyses of case-control data 30,31 . Dudbridge, Mahmoud and colleagues 32,33 proposed a correction for selection that occurs when performing case-only analyses. However, the same technique can correct for selection that is conditioned on any trait as long as GWAS can be performed on it.…”

Section: Discussionmentioning

confidence: 99%

Genetic analyses identify widespread sex-differential participation bias

Pirastu

Cordioli

Nandakumar

et al. 2020

Preprint

View full text Add to dashboard Cite

48Genetic association results are often interpreted with the assumption that study participation 49 does not affect downstream analyses. Understanding the genetic basis of this participation bias 50 is challenging as it requires the genotypes of unseen individuals. However, we demonstrate that 51 it is possible to estimate comparative biases by performing GWAS contrasting one subgroup 52 versus another. For example, we show that sex exhibits autosomal heritability in the presence of 53 sex-differential participation bias. By performing a GWAS of sex in ~3.3 million males and 54 females, we identify over 150 autosomal loci significantly associated with sex and highlight 55 complex traits underpinning differences in study participation between sexes. For example, the 56 body mass index (BMI) increasing allele at the FTO locus was observed at higher frequency in 57 males compared to females (OR 1.02 [1.02-1.03], P=4.4x10 -36 ). Finally, we demonstrate how 58 these biases can potentially lead to incorrect inferences in downstream analyses and propose a 59 conceptual framework for addressing such biases. Our findings highlight a new challenge that 60 genetic studies may face as sample sizes continue to grow. 61 62 63 64 65 66Individuals who enroll in research studies or purchase direct-to-consumer genetic tests are often 67 not representative of the general population 1,2,3 . 68For example, the UK Biobank study invited ~9 million individuals and achieved an overall 69 participation rate of 5.45% 4 . These enrolled individuals clearly demonstrated a "healthy 70 volunteer bias", with lower rates of obesity, smoking and fewer self-reported health conditions 71 than the sampling frame 4 . Achieving accurate representation of the sampling population in any 72 study is challenging. Examples do exist, however, such as the iPSYCH study which enrolled a 73 random sample of the population, based on DNA extracted from a nationwide collection of 74 neonatal dried blood spots 5 . The benefits of achieving such representativeness have long been 75 discussed 6,7,8,9 , with many arguing that unrepresentative samples can bias prevalence estimates 76 but do not necessarily create substantial biases on exposure-disease associations 10,11 . 77Purposely non-representative study designs can also be valuable, for example case-control 78 studies seeking to enrich cases with non-genetic risk factors can maximize power to detect 79 genetic effects 12 . 80 81Recent studies have highlighted that genetic factors are associated with aspects of study 82 engagement 13,14,15 . For example, individuals with high genetic risk for schizophrenia enrolled in 83 a study are less likely to complete health questionnaires, attend clinical assessments and 84 continue participation in longitudinal studies than those with lower genetic risk 13,16 . It remains 85 unclear to what extent genetic factors influence initial study participation, or what the 86 downstream consequences of such bias are, though there are prior attempts to quantify the bias 87 with simul...

show abstract

Section: Discussionmentioning

confidence: 99%

Genetic analyses identify widespread sex-differential participation bias

Pirastu

Cordioli

Nandakumar

et al. 2020

Preprint

View full text Add to dashboard Cite

show abstract

“…In primary care and trial settings, interventions often aim to improve these phenotypes rather than prevent their onset, which could have different causal relationships 13 . Methods are currently being developed which will enable us to investigate the causal effects of disease progression rather than disease onset, such as the recent slope hunter approach proposed by Mahmood and colleagues 41 .…”

Section: Budu-aggrey 21mentioning

confidence: 99%

Investigating the causal relationship between allergic disease and mental health

Budu-Aggrey

Joyce

Davies

et al. 2021

Preprint

View full text Add to dashboard Cite

Background Observational studies have reported an association between allergic disease and mental health, but a causal relationship has not been established. Objective To use Mendelian Randomization (MR) to investigate a possible causal relationship between atopic disease and mental health phenotypes. Methods The observational relationship between allergic disease and mental health was investigated in UK Biobank. The direction of causality was investigated with bidirectional two-sample MR using summary-level data from published genome-wide association studies. A genetic instrument was derived from associated variants for a broad allergic disease phenotype to test for causal relationships with various mental health outcomes. Genetic instruments were also derived for mental health conditions to assess causality in the reverse direction. We also investigated if these relationships were specific to atopic dermatitis (AD), asthma or hay fever. Results The broad allergic disease phenotype was phenotypically associated with most measures of mental health, but we found little evidence of causality in either direction. However, we did find evidence of genetic liability for bipolar disorder causally influencing hay fever risk (OR=0.94 per doubling odds of bipolar disorder risk; 95%CI=0.90-0.99; P-value=0.02), but evidence of a phenotypic association was weak. Conclusions Few of the phenotypic associations between allergic disease and mental health were replicated. Any causal effects we detected were considerably attenuated compared to the phenotypic association. This suggests that most co-morbidity observed clinically is unlikely to be causal.

show abstract

“…This is a major challenge affecting prognosis research. Several approaches are under development that aim to mitigate this type of bias, such as the "Slope-Hunter" method proposed by Mahmoud et al [43]. This method utilizes clustering technique to ultimately identify variants that only affect prognosis, and also to find estimated adjustment factors by identifying variants that affect incidence.…”

Section: Plos Onementioning

confidence: 99%

Genetic loci associated with prevalent and incident myocardial infarction and coronary heart disease in the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) Consortium

Hahn

Brown

et al. 2020

PLoS ONE

View full text Add to dashboard Cite

Background Genome-wide association studies have identified multiple genomic loci associated with coronary artery disease, but most are common variants in non-coding regions that provide limited information on causal genes and etiology of the disease. To overcome the limited scope that common variants provide, we focused our investigation on low-frequency and rare sequence variations primarily residing in coding regions of the genome. Methods and results Using samples of individuals of European ancestry from ten cohorts within the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) consortium, both cross-sectional and prospective analyses were conducted to examine associations between genetic variants and myocardial infarction (MI), coronary heart disease (CHD), and all-cause mortality following these events. For prevalent events, a total of 27,349 participants of European ancestry, including 1831 prevalent MI cases and 2518 prevalent CHD cases were used. For incident cases, a total of 55,736 participants of European ancestry were included (3,031 incident MI cases and 5,425 incident CHD cases). There were 1,860 all-cause deaths among the 3,751 MI and CHD cases from six cohorts that contributed to the analysis of all-cause mortality. Single variant and gene-based analyses were performed separately in each cohort and then meta-analyzed for each outcome. A low-frequency intronic variant (rs988583) in PLCL1 was significantly associated with prevalent MI (OR = 1.80, 95% confidence interval: 1.43, 2.27; P = 7.12 × 10−7). We conducted gene-based burden tests for genes with a cumulative minor allele count (cMAC) ≥ 5 and variants with minor allele frequency (MAF) < 5%. TMPRSS5 and LDLRAD1 were significantly associated with prevalent MI and CHD, respectively, and RC3H2 and ANGPTL4 were significantly associated with incident MI and CHD, respectively. No loci were significantly associated with all-cause mortality following a MI or CHD event. Conclusion This study identified one known locus (ANGPTL4) and four new loci (PLCL1, RC3H2, TMPRSS5, and LDLRAD1) associated with cardiovascular disease risk that warrant further investigation.

show abstract

Slope-Hunter: A robust method for index-event bias correction in genome-wide association studies of subsequent traits

Cited by 21 publications

References 23 publications

Genetic analyses identify widespread sex-differential participation bias

Genetic analyses identify widespread sex-differential participation bias

Investigating the causal relationship between allergic disease and mental health

Genetic loci associated with prevalent and incident myocardial infarction and coronary heart disease in the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) Consortium

Contact Info

Product

Resources

About