A unifying framework for robust association testing, estimation, and genetic model selection using the generalized linear model

Association of ZBTB38 gene polymorphism (rs724016) with height and fetal hemoglobin in individuals with sickle cell anemia

Objectives

Our study evaluated the association of the polymorphism rs724016 in the ZBTB38 gene, previously associated with height in other populations, with predictors of height, clinical outcomes, and laboratory parameters in sickle cell anemia (SCA).

Methods

Cross-sectional study with individuals with SCA and aged between 3 and 20 years. Clinical, laboratory, molecular, and bone age (BA) data were evaluated. Levels of IGF-1 and IGFBP-3 were adjusted for BA, target height (TH) was calculated as the mean parental height standard deviation score (SDS), and predicted adult height (PAH) SDS was calculated using BA.

Results

We evaluated 80 individuals with SCA. The homozygous genotype of the G allele of rs724016 was associated with a lower height SDS (p < 0.001) and, in a additive genetic model, was negatively associated with HbF levels (p = 0.016). Lower adjusted IGF-1 levels were associated with co-inheritance of alpha-thalassemia and with the absence of HU therapy. Elevated HbF levels were associated with a lower deficit in adjusted growth potential (TH minus PAH).

Conclusion

Our analysis shows that SNP rs724016 in the ZBTB38 is associated with shorter height and lower HbF levels, an important modifier of SCA.

Development and application of Single Primer Enrichment Technology (SPET) SNP assay for population genomics analysis and candidate gene discovery in lettuce

Single primer enrichment technology (SPET) is a novel high-throughput genotyping method based on short-read sequencing of specific genomic regions harboring polymorphisms. SPET provides an efficient and reproducible method for genotyping target loci, overcoming the limits associated with other reduced representation library sequencing methods that are based on a random sampling of genomic loci. The possibility to sequence regions surrounding a target SNP allows the discovery of thousands of closely linked, novel SNPs. In this work, we report the design and application of the first SPET panel in lettuce, consisting of 41,547 probes spanning the whole genome and designed to target both coding (~96%) and intergenic (~4%) regions. A total of 81,531 SNPs were surveyed in 160 lettuce accessions originating from a total of 10 countries in Europe, America, and Asia and representing 10 horticultural types. Model ancestry population structure clearly separated the cultivated accessions (Lactuca sativa) from accessions of its presumed wild progenitor (L. serriola), revealing a total of six genetic subgroups that reflected a differentiation based on cultivar typology. Phylogenetic relationships and principal component analysis revealed a clustering of butterhead types and a general differentiation between germplasm originating from Western and Eastern Europe. To determine the potentiality of SPET for gene discovery, we performed genome-wide association analysis for main agricultural traits in L. sativa using six models (GLM naive, MLM, MLMM, CMLM, FarmCPU, and BLINK) to compare their strength and power for association detection. Robust associations were detected for seed color on chromosome 7 at 50 Mbp. Colocalization of association signals was found for outer leaf color and leaf anthocyanin content on chromosome 9 at 152 Mbp and on chromosome 5 at 86 Mbp. The association for bolting time was detected with the GLM, BLINK, and FarmCPU models on chromosome 7 at 164 Mbp. Associations were detected in chromosomal regions previously reported to harbor candidate genes for these traits, thus confirming the effectiveness of SPET for GWAS. Our findings illustrated the strength of SPET for discovering thousands of variable sites toward the dissection of the genomic diversity of germplasm collections, thus allowing a better characterization of lettuce collections.

Genome-wide association testing in malaria studies in the presence of overdominance

BACKGROUND

In human genetics, heterozygote advantage (heterosis) has been detected in studies that focused on specific genes but not in genome-wide association studies (GWAS). For example, heterosis is believed to confer resistance to certain strains of malaria in patients heterozygous for the sickle-cell gene, haemoglobin S (HbS). Yet the power of allelic tests can be substantially diminished by heterosis. Since GWAS (and haplotype-associations) also utilize allelic tests, it is unclear to what degree GWAS could underachieve because heterosis is ignored.

METHODS

In this study, a two-step approach to genetic association testing in malaria studies in a GWAS setting that may enhance the power of the tests was proposed, by identifying the underlying genetic model first before applying the association tests. Generalized linear models for dominant, recessive, additive, and heterotic effects were fitted and model selection was performed. This was achieved via tests of significance using the MAX and allelic tests, noting the minimum p-values across all the models and the proportion of tests that a given genetic model was deemed the best. An example dataset, based on 17 SNPs, from a robust genetic association study and simulated genotype datasets, were used to illustrate the method. Case-control genotype data on malaria from Kenya and Gambia were used for validation.

RESULTS AND CONCLUSION

Results showed that the allelic test returned some false negatives under the heterosis model, suggesting reduced power in testing genetic association. Disparities were observed for some chromosomes in the Kenyan and Gambian datasets, including the sex chromosomes. Thus, GWAS and haplotype associations should be treated with caution, unless the underlying genetic model had been determined.

A nonparametric alternative to the Cochran-Armitage trend test in genetic case-control association studies: The Jonckheere-Terpstra trend test

Identifications of novel genetic signals conferring susceptibility to human complex diseases is pivotal to the disease diagnosis, prevention, and treatment. Genetic association study is a powerful tool to discover candidate genetic signals that contribute to diseases, through statistical tests for correlation between the disease status and genetic variations in study samples. In such studies with a case-control design, a standard practice is to perform the Cochran-Armitage (CA) trend test under an additive genetic model, which suffers from power loss when the model assumption is wrong. The Jonckheere-Terpstra (JT) trend test is an alternative method to evaluate association in a nonparametric way. This study compares the power of the JT trend test and the CA trend test in various scenarios, including different sample sizes (200-2000), minor allele frequencies (0.05-0.4), and underlying modes of inheritance (dominant genetic model to recessive genetic model). By simulation and real data analysis, it is shown that in general the JT trend test has higher, similar, and lower power than the CA trend test when the underlying mode of inheritance is dominant, additive, and recessive, respectively; when the sample size is small and the minor allele frequency is low, the JT trend test outperforms the CA trend test across the spectrum of genetic models. In sum, the JT trend test is a valuable alternative to the CA trend test under certain circumstances with higher statistical power, which could lead to better detection of genetic signals to human diseases and finer dissection of their genetic architecture.

General regression model: A "model-free" association test for quantitative traits allowing to test for the underlying genetic model

Most genome-wide association studies used genetic-model-based tests assuming an additive mode of inheritance, leading to underpowered association tests in case of departure from additivity. The general regression model (GRM) association test proposed by Fisher and Wilson in 1980 makes no assumption on the genetic model. Interestingly, it also allows formal testing of the underlying genetic model. We conducted a simulation study of quantitative traits to compare the power of the GRM test to the classical linear regression tests, the maximum of the three statistics (MAX), and the allele-based (allelic) tests. Simulations were performed on two samples sizes, using a large panel of genetic models, varying genetic models, minor allele frequencies, and the percentage of explained variance. In case of departure from additivity, the GRM was more powerful than the additive regression tests (power gain reaching 80%) and had similar power when the true model is additive. GRM was also as or more powerful than the MAX or allelic tests. The true simulated model was mostly retained by the GRM test. Application of GRM to HbA1c illustrates its gain in power. To conclude, GRM increases power to detect association for quantitative traits, allows determining the genetic model and is easily applicable.

An efficient integrative resampling method for gene-trait association analysis

Genetic association studies are popular for identifying genetic variants, such as single nucleotide polymorphisms (SNPs), that are associated with complex traits. Statistical tests are commonly performed one SNP at a time with an assumed mode of inheritance such as recessive, additive, or dominant genetic model. Such analysis can result in inadequate power when the employed model deviates from the underlying true genetic model. We propose an integrative association test procedure under a generalized linear model framework to flexibly model the data from the above three common genetic models and beyond. A computationally efficient resampling procedure is adopted to estimate the null distribution of the proposed test statistic. Simulation results show that our methods maintain the Type I error rate irrespective of the existence of confounding covariates and achieve adequate power compared to the methods with the true genetic model. The new methods are applied to two genetic studies on the resistance of severe malaria and sarcoidosis.

Common variants in the CLDN2-MORC4 and PRSS1-PRSS2 loci confer susceptibility to acute pancreatitis

BACKGROUND/OBJECTIVES

Acute pancreatitis (AP) is one of the most common gastrointestinal disorders often requiring hospitalization. Frequent aetiologies are gallstones and alcohol abuse. In contrast to chronic pancreatitis (CP) few robust genetic associations have been described. Here we analysed whether common variants in the CLDN2-MORC4 and the PRSS1-PRSS2 locus that increase recurrent AP and CP risk associate with AP.

METHODS

We screened 1462 AP patients and 3999 controls with melting curve analysis for SNPs rs10273639 (PRSS1-PRSS2), rs7057398 (RIPPLY), and rs12688220 (MORC4). Calculations were performed for the overall group, aetiology, and gender sub-groups. To examine genotype-phenotype relationships we performed several meta-analyses.

RESULTS

Meta-analyses of all AP patients depicted significant (p-value < 0.05) associations for rs10273639 (odds ratio (OR) 0.88, 95% confidence interval (CI) 0.81-0.97, p-value 0.01), rs7057398 (OR 1.27, 95% CI 1.07-1.5, p-value 0.005), and rs12688220 (OR 1.32, 95% CI 1.12-1.56, p-value 0.001). For the different aetiology groups a significant association was shown for rs10273639 (OR 0.76, 95% CI 0.63-0.92, p-value 0.005), rs7057398 (OR 1.43, 95% CI 1.07-1.92, p-value 0.02), and rs12688220 (OR 1.44, 95% CI 1.07-1.93, p-value 0.02) in the alcoholic sub-group only.

CONCLUSIONS

The association of CP risk variants with different AP aetiologies, which is strongest in the alcoholic AP group, might implicate common pathomechanisms most likely between alcoholic AP and CP.

Decomposing Pearson's χ2 test: A linear regression and its departure from linearity

In case-control genetic association studies, a standard practice is to perform the Cochran-Armitage (CA) trend test under the assumption of the additive model because of its robustness. We could even identify situations in which it outperformed the analysis model consistent with the underlying inheritance mode. In this article, we analytically reveal the statistical basis that leads to the phenomenon. By elucidating the origin of the CA trend test as a linear regression model, we decompose Pearson's χ² -test statistic into two components-one is the CA trend test statistic that measures the goodness of fit of the linear regression model, and the other measures the discrepancy between data and the linear regression model. Under this framework, we show that the additive coding scheme, as well as the multiplicative coding scheme, increases the coefficient of determination of the regression model by increasing the spread of data points. We also obtain the conditions under which the CA trend test statistic equals the MAX statistic and Pearson's χ² -test statistic.

Genetic model misspecification in genetic association studies

Objective

The underlying model of the genetic determinant of a trait is generally not known with certainty a priori. Hence, in genetic association studies, a dominant model might be erroneously modelled as additive, an error investigated previously. We explored this question, for candidate gene studies, by evaluating the sample size required to compensate for the misspecification and improve inference at the analysis stage. Power calculations were carried out with (1) the true dominant model and (2) the incorrect additive model. Empirical power, sample size and effect size were compared between scenarios (1) and (2). In each of the scenarios the estimates were evaluated for a rare (minor allele frequency < 0.01), low frequency (0.01 ≤ minor allele frequency < 0.05) and common (minor allele frequency ≥ 0.05) single nucleotide polymorphism.

Results

The results confirm the detrimental effect of the misspecification error on power and effect size for any minor allele frequency. The implications of the error are not negligible; therefore, candidate gene studies should consider the more conservative sample size to compensate for the effect of error. When it is not possible to extend the sample size, methods that help mitigate the impact of the error should be systematically used.

Electronic supplementary material

The online version of this article (10.1186/s13104-017-2911-3) contains supplementary material, which is available to authorized users.

Differentiating the Cochran-Armitage Trend Test and Pearson's χ2 Test: Location and Dispersion

In genetic case-control association studies, a standard practice is to perform the Cochran-Armitage (CA) trend test with 1 degree-of-freedom (d.f.) under the assumption of an additive model. However, when the true genetic model is recessive or near recessive, it is outperformed by Pearson's χ² test with 2 d.f. In this article, we analytically reveal the statistical basis that leads to the phenomenon. First, we show that the CA trend test examines the location shift between the case and control groups, whereas Pearson's χ² test examines both the location and dispersion shifts between the two groups. Second, we show that under the additive model, the effect of location deviation outweighs that of the dispersion deviation and vice versa under a near recessive model. Therefore, Pearson's χ² test is a more robust test than the CA trend test, and it outperforms the latter when the mode of inheritance evolves to the recessive end.

Gain of power of the general regression model compared to Cochran-Armitage Trend tests: simulation study and application to bipolar disorder

Background

Most genome-wide association studies assumed an additive model of inheritance which may result in significant loss of power when there is a strong departure from additivity. The General Regression Model (GRM), which allows performing an assumption-free test for association by testing for both additive effect and deviation from additive effect, may be more appropriate for association tests. Additionally, GRM allows testing the underlying genetic model. We compared the power of GRM association test to additive and other Cochran-Armitage Trend (CAT) tests through simulations and by applying GRM to a large case/control sample, the bipolar Welcome Trust Case Control Cohort data. Simulations were performed on two sets of case/control samples (1000/1000 and 2000/2000), using a large panel of genetic models. Four association tests (GRM and additive, recessive and dominant CAT tests) were applied to all replicates.

Results

We showed that GRM power to detect association was similar or greater than the additive CAT test, in particular in case of recessive inheritance, with up to 67% gain in power. GRM analysis of genome-wide bipolar disorder Welcome Trust Consortium data (1998 cases/3004 controls) showed significant association in the 16p12 region (rs420259; P = 3.4E-7) which has not been identified using the additive CAT test. As expected, rs42025 fitted a non-additive (recessive) model.

Conclusions

GRM provides increased power compared to the additive CAT test for association studies and is easily applicable.

Electronic supplementary material

The online version of this article (doi:10.1186/s12863-017-0486-6) contains supplementary material, which is available to authorized users.

PNPLA3 rs738409 and TM6SF2 rs58542926 variants increase the risk of hepatocellular carcinoma in alcoholic cirrhosis

BACKGROUND

PNPLA3 rs738409 polymorphism is associated with fatty liver disease, alcoholic or non-alcoholic (NAFLD) and hepatocellular carcinoma (HCC). TM6SF2 rs58542926 is clearly associated with NAFLD, but it is not clearly associated with HCC. The relationship between TM6SF2 rs58542926 and HCC and the potential synergistic effect of TM6SF2 and PNPLA3 variants in modifying the risk of HCC are not known.

AIM

This study assessed the interaction between PNPLA3 rs738409 and TM6SF2 rs58542926 variants in the conditioning of HCC development.

METHODS

A total of 511 cirrhotic patients (44% alcohol-related, 56% viral, 57.5% liver transplanted) were retrospectively investigated for HCC occurrence. PNPLA3 rs734809 and TM6SF2 rs58542926 were genotyped using restriction fragment length polymorphism and real-time allelic discrimination polymerase chain reaction methods.

RESULTS

Patients with HCC were more likely to be PNPLA3 rs734809 G/G homozygotes (41/150 vs. 60/361, p=0.009) or TM6SF2 rs58542926 C/T-T/T (27/150 vs. 41/361, p=0.044). The presence of either PNPLA3 G/G or TM6SF2*/T identified high-risk genotypes for HCC, which were strongly associated with HCC (64/150 vs. 93/361, p=0.0002). This association was evident in alcohol-related (p=0.0007) but not in viral cirrhosis.

CONCLUSION

TM6SF2 C/T or T/T in conjunction with PNPLA3 G/G variants may be potential genetic risk factors for developing HCC in alcohol-related cirrhosis.

Polymorphisms at PRSS1-PRSS2 and CLDN2-MORC4 loci associate with alcoholic and non-alcoholic chronic pancreatitis in a European replication study

OBJECTIVE

Several genetic risk factors have been identified for non-alcoholic chronic pancreatitis (NACP). A genome-wide association study reported an association of chronic pancreatitis (CP) with variants in PRSS1-PRSS2 (rs10273639; near the gene encoding cationic trypsinogen) and CLDN2-MORC4 loci (rs7057398 in RIPPLY1 and rs12688220 in MORC4). We aimed to refine these findings in a large European cohort.

DESIGN

We studied 3062 patients with alcohol-related CP (ACP) or NACP and 5107 controls. Also, 1559 German patients with alcohol-associated cirrhosis or alcohol dependence were included for comparison. We performed several meta-analyses to examine genotype-phenotype relationships.

RESULTS

Association with ACP was found for rs10273639 (OR, 0.63; 95% CI 0.55 to 0.72). ACP was also associated with variants rs7057398 and rs12688220 in men (OR, 2.26; 95% CI 1.94 to 2.63 and OR, 2.66; 95% CI 2.21 to 3.21, respectively) and in women (OR, 1.57; 95% CI 1.14 to 2.18 and OR 1.71; 95% CI 1.41 to 2.07, respectively). Similar results were obtained when German patients with ACP were compared with those with alcohol-associated cirrhosis or alcohol dependence. In the overall population of patients with NACP, association with rs10273639 was absent (OR, 0.93; 95% CI 0.79 to 1.01), whereas rs7057398 of the X chromosomal single nucleotide polymorphisms was associated with NACP in women only (OR, 1.32; 95% CI 1.15 to 1.51).

CONCLUSIONS

The single-nucleotide polymorphisms rs10273639 at the PRSS1-PRSS2 locus and rs7057398 and rs12688220 at the CLDN2-MORC4 locus are associated with CP and strongly associate with ACP, but only rs7057398 with NACP in female patients.

Nonadditive Effects of Genes in Human Metabolomics

Genome-wide association studies (GWAS) are widely applied to analyze the genetic effects on phenotypes. With the availability of high-throughput technologies for metabolite measurements, GWAS successfully identified loci that affect metabolite concentrations and underlying pathways. In most GWAS, the effect of each SNP on the phenotype is assumed to be additive. Other genetic models such as recessive, dominant, or overdominant were considered only by very few studies. In contrast to this, there are theories that emphasize the relevance of nonadditive effects as a consequence of physiologic mechanisms. This might be especially important for metabolites because these intermediate phenotypes are closer to the underlying pathways than other traits or diseases. In this study we analyzed systematically nonadditive effects on a large panel of serum metabolites and all possible ratios (22,801 total) in a population-based study [Cooperative Health Research in the Region of Augsburg (KORA) F4, N = 1,785]. We applied four different 1-degree-of-freedom (1-df) tests corresponding to an additive, dominant, recessive, and overdominant trait model as well as a genotypic model with two degree-of-freedom (2-df) that allows a more general consideration of genetic effects. Twenty-three loci were found to be genome-wide significantly associated (Bonferroni corrected P ≤ 2.19 × 10(-12)) with at least one metabolite or ratio. For five of them, we show the evidence of nonadditive effects. We replicated 17 loci, including 3 loci with nonadditive effects, in an independent study (TwinsUK, N = 846). In conclusion, we found that most genetic effects on metabolite concentrations and ratios were indeed additive, which verifies the practice of using the additive model for analyzing SNP effects on metabolites.