Predicting Disease Risk Using Bootstrap Ranking and Classification Algorithms

Genome-wide association studies (GWAS) are widely used to search for genetic loci that underlie human disease. Another goal is to predict disease risk for different individuals given their genetic sequence. Such predictions could either be used as a “black box” in order to promote changes in life-style and screening for early diagnosis, or as a model that can be studied to better understand the mechanism of the disease. Current methods for risk prediction typically rank single nucleotide polymorphisms (SNPs) by the p-value of their association with the disease, and use the top-associated SNPs as input to a classification algorithm. However, the predictive power of such methods is relatively poor. To improve the predictive power, we devised BootRank, which uses bootstrapping in order to obtain a robust prioritization of SNPs for use in predictive models. We show that BootRank improves the ability to predict disease risk of unseen individuals in the Wellcome Trust Case Control Consortium (WTCCC) data and results in a more robust set of SNPs and a larger number of enriched pathways being associated with the different diseases. Finally, we show that combining BootRank with seven different classification algorithms improves performance compared to previous studies that used the WTCCC data. Notably, diseases for which BootRank results in the largest improvements were recently shown to have more heritability than previously thought, likely due to contributions from variants with low minimum allele frequency (MAF), suggesting that BootRank can be beneficial in cases where SNPs affecting the disease are poorly tagged or have low MAF. Overall, our results show that improving disease risk prediction from genotypic information may be a tangible goal, with potential implications for personalized disease screening and treatment.     

Characterization of SNPs Associated with Prostate Cancer in Men of Ashkenazic Descent from the Set of GWAS Identified SNPs: Impact of Cancer Family History and Cumulative SNP Risk Prediction

Background

Genome-wide association studies (GWAS) have identified multiple SNPs associated with prostate cancer (PrCa). Population isolates may have different sets of risk alleles for PrCa constituting unique population and individual risk profiles.

Methods

To test this hypothesis, associations between 31 GWAS SNPs of PrCa were examined among 979 PrCa cases and 1,251 controls of Ashkenazic descent using logistic regression. We also investigated risks by age at diagnosis, pathological features of PrCa, and family history of cancer. Moreover, we examined associations between cumulative number of risk alleles and PrCa and assessed the utility of risk alleles in PrCa risk prediction by comparing the area under the curve (AUC) for different logistic models.

Results

Of the 31 genotyped SNPs, 8 were associated with PrCa at p≤0.002 (corrected p-value threshold) with odds ratios (ORs) ranging from 1.22 to 1.42 per risk allele. Four SNPs were associated with aggressive PrCa, while three other SNPs showed potential interactions for PrCa by family history of PrCa (rs8102476; 19q13), lung cancer (rs17021918; 4q22), and breast cancer (rs10896449; 11q13). Men in the highest vs. lowest quartile of cumulative number of risk alleles had ORs of 3.70 (95% CI 2.76–4.97); 3.76 (95% CI 2.57–5.50), and 5.20 (95% CI 2.94–9.19) for overall PrCa, aggressive cancer and younger age at diagnosis, respectively. The addition of cumulative risk alleles to the model containing age at diagnosis and family history of PrCa yielded a slightly higher AUC (0.69 vs. 0.64).

Conclusion

These data define a set of risk alleles associated with PrCa in men of Ashkenazic descent and indicate possible genetic differences for PrCa between populations of European and Ashkenazic ancestry. Use of genetic markers might provide an opportunity to identify men at highest risk for younger age of onset PrCa; however, their clinical utility in identifying men at highest risk for aggressive cancer remains limited.     

MicroRNA Related Polymorphisms and Breast Cancer Risk

Genetic variations, such as single nucleotide polymorphisms (SNPs) in microRNAs (miRNA) or in the miRNA binding sites may affect the miRNA dependent gene expression regulation, which has been implicated in various cancers, including breast cancer, and may alter individual susceptibility to cancer. We investigated associations between miRNA related SNPs and breast cancer risk. First we evaluated 2,196 SNPs in a case-control study combining nine genome wide association studies (GWAS). Second, we further investigated 42 SNPs with suggestive evidence for association using 41,785 cases and 41,880 controls from 41 studies included in the Breast Cancer Association Consortium (BCAC). Combining the GWAS and BCAC data within a meta-analysis, we estimated main effects on breast cancer risk as well as risks for estrogen receptor (ER) and age defined subgroups. Five miRNA binding site SNPs associated significantly with breast cancer risk: rs1045494 (odds ratio (OR) 0.92; 95% confidence interval (CI): 0.88–0.96), rs1052532 (OR 0.97; 95% CI: 0.95–0.99), rs10719 (OR 0.97; 95% CI: 0.94–0.99), rs4687554 (OR 0.97; 95% CI: 0.95–0.99, and rs3134615 (OR 1.03; 95% CI: 1.01–1.05) located in the 3′ UTR of CASP8, HDDC3, DROSHA, MUSTN1, and MYCL1, respectively. DROSHA belongs to miRNA machinery genes and has a central role in initial miRNA processing. The remaining genes are involved in different molecular functions, including apoptosis and gene expression regulation. Further studies are warranted to elucidate whether the miRNA binding site SNPs are the causative variants for the observed risk effects.     

A genome-wide screen of gene–gene interactions for rheumatoid arthritis susceptibility

The objective of the study was to identify interacting genes contributing to rheumatoid arthritis (RA) susceptibility and identify SNPs that discriminate between RA patients who were anti-cyclic citrullinated protein positive and healthy controls. We analyzed two independent cohorts from the North American Rheumatoid Arthritis Consortium. A cohort of 908 RA cases and 1,260 controls was used to discover pairwise interactions among SNPs and to identify a set of single nucleotide polymorphisms (SNPs) that predict RA status, and a second cohort of 952 cases and 1,760 controls was used to validate the findings. After adjusting for HLA-shared epitope alleles, we identified and replicated seven SNP pairs within the HLA class II locus with significant interaction effects. We failed to replicate significant pairwise interactions among non-HLA SNPs. The machine learning approach “random forest” applied to a set of SNPs selected from single-SNP and pairwise interaction tests identified 93 SNPs that distinguish RA cases from controls with 70% accuracy. HLA SNPs provide the most classification information, and inclusion of non-HLA SNPs improved classification. While specific gene–gene interactions are difficult to validate using genome-wide SNP data, a stepwise approach combining association and classification methods identifies candidate interacting SNPs that distinguish RA cases from healthy controls.     

Combination Testing Using a Single MSH5 Variant alongside HLA Haplotypes Improves the Sensitivity of Predicting Coeliac Disease Risk in the Polish Population

Assessment of non-HLA variants alongside standard HLA testing was previously shown to improve the identification of potential coeliac disease (CD) patients. We intended to identify new genetic variants associated with CD in the Polish population that would improve CD risk prediction when used alongside HLA haplotype analysis. DNA samples of 336 CD and 264 unrelated healthy controls were used to create DNA pools for a genome wide association study (GWAS). GWAS findings were validated with individual HLA tag single nucleotide polymorphism (SNP) typing of 473 patients and 714 healthy controls. Association analysis using four HLA-tagging SNPs showed that, as was found in other populations, positive predicting genotypes (HLA-DQ2.5/DQ2.5, HLA-DQ2.5/DQ2.2, and HLA-DQ2.5/DQ8) were found at higher frequencies in CD patients than in healthy control individuals in the Polish population. Both CD-associated SNPs discovered by GWAS were found in the CD susceptibility region, confirming the previously-determined association of the major histocompatibility (MHC) region with CD pathogenesis. The two most significant SNPs from the GWAS were rs9272346 (HLA-dependent; localized within 1 Kb of DQA1) and rs3130484 (HLA-independent; mapped to MSH5). Specificity of CD prediction using the four HLA-tagging SNPs achieved 92.9%, but sensitivity was only 45.5%. However, when a testing combination of the HLA-tagging SNPs and the MSH5 SNP was used, specificity decreased to 80%, and sensitivity increased to 74%. This study confirmed that improvement of CD risk prediction sensitivity could be achieved by including non-HLA SNPs alongside HLA SNPs in genetic testing.     

Distinct SNP Combinations Confer Susceptibility to Urinary Bladder Cancer in Smokers and Non-Smokers

Recently, genome-wide association studies have identified and validated genetic variations associated with urinary bladder cancer (UBC). However, it is still unknown whether the high-risk alleles of several SNPs interact with one another, leading to an even higher disease risk. Additionally, there is no information available on how the UBC risk due to these SNPs compare to the risk of cigarette smoking and to occupational exposure to urinary bladder carcinogens, and whether the same or different SNP combinations are relevant in smokers and non-smokers. To address these questions, we analyzed the genotypes of six SNPs, previously found to be associated with UBC, together with the GSTM1 deletion, in 1,595 UBC cases and 1,760 controls, stratified for smoking habits. We identified the strongest interactions of different orders and tested the stability of their effect by bootstrapping. We found that different SNP combinations were relevant in smokers and non-smokers. In smokers, polymorphisms involved in detoxification of cigarette smoke carcinogens were most relevant (GSTM1, rs11892031), in contrast to those in non-smokers with MYC and APOBEC3A near polymorphisms (rs9642880, rs1014971) being the most influential. Stable combinations of up to three high-risk alleles resulted in higher odds ratios (OR) than the individual SNPs, although the interaction effect was less than additive. The highest stable combination effects resulted in an OR of about 2.0, which is still lower than the ORs of cigarette smoking (here, current smokers'' OR: 3.28) and comparable to occupational carcinogen exposure risks which, depending on the workplace, show mostly ORs up to 2.0.     

Evaluation of Parkinson Disease Risk Variants as Expression-QTLs

The recent Parkinson Disease GWAS Consortium meta-analysis and replication study reports association at several previously confirmed risk loci SNCA, MAPT, GAK/DGKQ, and HLA and identified a novel risk locus at RIT2. To further explore functional consequences of these associations, we investigated modification of gene expression in prefrontal cortex brain samples of pathologically confirmed PD cases (N = 26) and controls (N = 24) by 67 associated SNPs in these 5 loci. Association between the eSNPs and expression was evaluated using a 2-degrees of freedom test of both association and difference in association between cases and controls, adjusted for relevant covariates. SNPs at each of the 5 loci were tested for cis-acting effects on all probes within 250 kb of each locus. Trans-effects of the SNPs on the 39,122 probes passing all QC on the microarray were also examined. From the analysis of cis-acting SNP effects, several SNPs in the MAPT region show significant association to multiple nearby probes, including two strongly correlated probes targeting the gene LOC644246 and the duplicated genes LRRC37A and LRRC37A2, and a third uncorrelated probe targeting the gene DCAKD. Significant cis-associations were also observed between SNPs and two probes targeting genes in the HLA region on chromosome 6. Expanding the association study to examine trans effects revealed an additional 23 SNP-probe associations reaching statistical significance (p<2.8×10⁻⁸) including SNPs from the SNCA, MAPT and RIT2 regions. These findings provide additional context for the interpretation of PD associated SNPs identified in recent GWAS as well as potential insight into the mechanisms underlying the observed SNP associations.     

Investigation of Caucasian rheumatoid arthritis susceptibility loci in African patients with the same disease

Introduction

The largest genetic risk to develop rheumatoid arthritis (RA) arises from a group of alleles of the HLA DRB1 locus ('shared epitope', SE). Over 30 non-HLA single nucleotide polymorphisms (SNPs) predisposing to disease have been identified in Caucasians, but they have never been investigated in West/Central Africa. We previously reported a lower prevalence of the SE in RA patients in Cameroon compared to European patients and aimed in the present study to investigate the contribution of Caucasian non-HLA RA SNPs to disease susceptibility in Black Africans.

Methods

RA cases and controls from Cameroon were genotyped for Caucasian RA susceptibility SNPs using Sequenom MassArray technology. Genotype data were also available for 5024 UK cases and 4281 UK controls and for 119 Yoruba individuals in Ibadan, Nigeria (YRI, HapMap). A Caucasian aggregate genetic-risk score (GRS) was calculated as the sum of the weighted risk-allele counts.

Results

After genotyping quality control procedures were performed, data on 28 Caucasian non-HLA susceptibility SNPs were available in 43 Cameroonian RA cases and 44 controls. The minor allele frequencies (MAF) were tightly correlated between Cameroonian controls and YRI individuals (correlation coefficient 93.8%, p = 1.7E-13), and they were pooled together. There was no correlation between MAF of UK and African controls; 13 markers differed by more than 20%. The MAF for markers at PTPN22, IL2RA, FCGR2A and IL2/IL21 was below 2% in Africans. The GRS showed a strong association with RA in the UK. However, the GRS did not predict RA in Africans (OR = 0.71, 95% CI 0.29 - 1.74, p = 0.456). Random sampling from the UK cohort showed that this difference in association is unlikely to be explained by small sample size or chance, but is statistically significant with p<0.001.

Conclusions

The MAFs of non-HLA Caucasian RA susceptibility SNPs are different between Caucasians and Africans, and several polymorphisms are barely detectable in West/Central Africa. The genetic risk of developing RA conferred by a set of 28 Caucasian susceptibility SNPs is significantly different between the UK and Africa with p<0.001. Taken together, these observations strengthen the hypothesis that the genetic architecture of RA susceptibility is different in different ethnic backgrounds.     

Effective Genetic-Risk Prediction Using Mixed Models

For predicting genetic risk, we propose a statistical approach that is specifically adapted to dealing with the challenges imposed by disease phenotypes and case-control sampling. Our approach (termed Genetic Risk Scores Inference [GeRSI]), combines the power of fixed-effects models (which estimate and aggregate the effects of single SNPs) and random-effects models (which rely primarily on whole-genome similarities between individuals) within the framework of the widely used liability-threshold model. We demonstrate in extensive simulation that GeRSI produces predictions that are consistently superior to current state-of-the-art approaches. When applying GeRSI to seven phenotypes from the Wellcome Trust Case Control Consortium (WTCCC) study, we confirm that the use of random effects is most beneficial for diseases that are known to be highly polygenic: hypertension (HT) and bipolar disorder (BD). For HT, there are no significant associations in the WTCCC data. The fixed-effects model yields an area under the ROC curve (AUC) of 54%, whereas GeRSI improves it to 59%. For BD, using GeRSI improves the AUC from 55% to 62%. For individuals ranked at the top 10% of BD risk predictions, using GeRSI substantially increases the BD relative risk from 1.4 to 2.5.     

Age at onset determines severity and choice of treatment in early rheumatoid arthritis: a prospective study

Introduction

Disease activity, severity and comorbidity contribute to increased mortality in patients with rheumatoid arthritis (RA). We evaluated the impact of age at disease onset on prognostic risk factors and treatment in patients with early disease.

Methods

In this study, 950 RA patients were followed regularly from the time of inclusion (<12 months from symptom onset) for disease activity (erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), tender and/or swollen joints, Visual Analogue Scale pain and global scores, and Disease Activity Score in 28 joints (DAS28)) and function (Health Assessment Questionnaire (HAQ)). Disease severity, measured on the basis of radiographs of the hands and feet (erosions based on Larsen score), extraarticular disease, nodules, and comorbidities and treatment (disease-modifying antirheumatic drugs (DMARDs), corticosteroids, biologics and nonsteroidal anti-inflammatory drugs) were recorded at the time of inclusion and at 5 years. Autoantibodies (rheumatoid factor, antinuclear antibodies and antibodies against cyclic citrullinated peptides (ACPAs)) and genetic markers (human leucocyte antibody (HLA) shared epitope and protein tyrosine phosphatase nonreceptor type 22 (PTPN22)) were analysed at the time of inclusion. Data were stratified as young-onset RA (YORA) and late-onset RA (LORA), which were defined as being below or above the median age at the time of onset of RA (58 years).

Results

LORA was associated with lower frequency of ACPA (P < 0.05) and carriage of PTPN22-T variant (P < 0.01), but with greater disease activity at the time of inclusion measured on the basis of ESR (P < 0.001), CRP (P < 0.01) and accumulated disease activity (area under the curve for DAS28 score) at 6 months (P < 0.01), 12 months (P < 0.01) and 24 months (P < 0.05), as well as a higher HAQ score (P < 0.01) compared with YORA patients. At baseline and 24 months, LORA was more often associated with erosions (P < 0.01 for both) and higher Larsen scores (P < 0.001 for both). LORA was more often treated with corticosteroids (P < 0.01) and less often with methotrexate (P < 0.001) and biologics (P < 0.001). YORA was more often associated with early DMARD treatment (P < 0.001). The results of multiple regression analyses supported our findings regarding the impact of age on chosen treatment.

Conclusion

YORA patients were more frequently ACPA-positive than LORA patients. LORA was more often associated with erosions, higher Larsen scores, higher disease activity and higher HAQ scores at baseline. Nevertheless, YORA was treated earlier with DMARDs, whilst LORA was more often treated with corticosteroids and less often with DMARDs in early-stage disease. These findings could have implications for the development of comorbidities.     

Association of cigarette smoking and CRP levels with DNA methylation in α-1 antitrypsin deficiency

Alpha-1 antitrypsin (AAT) deficiency and tobacco smoking are confirmed risk factors for Chronic Obstructive Pulmonary Disease. We hypothesized that variable DNA methylation would be associated with smoking and inflammation, as reflected by the level of C-Reactive Protein (CRP) in AAT-deficient subjects. Methylation levels of 1,411 autosomal CpG sites from the Illumina GoldenGate Methylation Cancer Panel I were analyzed in 316 subjects. Associations of five smoking behaviors and CRP levels with individual CpG sites and average methylation levels were assessed using non-parametric testing, linear regression and linear mixed effect models, with and without adjustment for age and gender. Univariate linear regression analysis revealed that methylation levels of 16 CpG sites significantly associated with ever-smoking status. A CpG site in the TGFBI gene was the only site associated with ever-smoking after adjustment for age and gender. No highly significant associations existed between age at smoking initiation, pack-years smoked, duration of smoking, and time since quitting smoking as predictors of individual CpG site methylation levels. However, ever-smoking and younger age at smoking initiation associated with lower methylation level averaged across all sites. DNA methylation at CpG sites in the RUNX3, JAK3 and KRT1 genes associated with CRP levels. The most significantly associated CpG sites with gender and age mapped to the CASP6 and FZD9 genes, respectively. In summary, this study identified multiple potential candidate CpG sites associated with ever-smoking and CRP level in AAT-deficient subjects. Phenotypic variability in Mendelian diseases may be due to epigenetic factors.     

Simultaneous Discovery,Estimation and Prediction Analysis of Complex Traits Using a Bayesian Mixture Model

Gene discovery, estimation of heritability captured by SNP arrays, inference on genetic architecture and prediction analyses of complex traits are usually performed using different statistical models and methods, leading to inefficiency and loss of power. Here we use a Bayesian mixture model that simultaneously allows variant discovery, estimation of genetic variance explained by all variants and prediction of unobserved phenotypes in new samples. We apply the method to simulated data of quantitative traits and Welcome Trust Case Control Consortium (WTCCC) data on disease and show that it provides accurate estimates of SNP-based heritability, produces unbiased estimators of risk in new samples, and that it can estimate genetic architecture by partitioning variation across hundreds to thousands of SNPs. We estimated that, depending on the trait, 2,633 to 9,411 SNPs explain all of the SNP-based heritability in the WTCCC diseases. The majority of those SNPs (>96%) had small effects, confirming a substantial polygenic component to common diseases. The proportion of the SNP-based variance explained by large effects (each SNP explaining 1% of the variance) varied markedly between diseases, ranging from almost zero for bipolar disorder to 72% for type 1 diabetes. Prediction analyses demonstrate that for diseases with major loci, such as type 1 diabetes and rheumatoid arthritis, Bayesian methods outperform profile scoring or mixed model approaches.     

Pathway Analysis of GWAS Provides New Insights into Genetic Susceptibility to 3 Inflammatory Diseases

Although the introduction of genome-wide association studies (GWAS) have greatly increased the number of genes associated with common diseases, only a small proportion of the predicted genetic contribution has so far been elucidated. Studying the cumulative variation of polymorphisms in multiple genes acting in functional pathways may provide a complementary approach to the more common single SNP association approach in understanding genetic determinants of common disease. We developed a novel pathway-based method to assess the combined contribution of multiple genetic variants acting within canonical biological pathways and applied it to data from 14,000 UK individuals with 7 common diseases. We tested inflammatory pathways for association with Crohn''s disease (CD), rheumatoid arthritis (RA) and type 1 diabetes (T1D) with 4 non-inflammatory diseases as controls. Using a variable selection algorithm, we identified variants responsible for the pathway association and evaluated their use for disease prediction using a 10 fold cross-validation framework in order to calculate out-of-sample area under the Receiver Operating Curve (AUC). The generalisability of these predictive models was tested on an independent birth cohort from Northern Finland. Multiple canonical inflammatory pathways showed highly significant associations (p 10⁻³–10⁻²⁰) with CD, T1D and RA. Variable selection identified on average a set of 205 SNPs (149 genes) for T1D, 350 SNPs (189 genes) for RA and 493 SNPs (277 genes) for CD. The pattern of polymorphisms at these SNPS were found to be highly predictive of T1D (91% AUC) and RA (85% AUC), and weakly predictive of CD (60% AUC). The predictive ability of the T1D model (without any parameter refitting) had good predictive ability (79% AUC) in the Finnish cohort. Our analysis suggests that genetic contribution to common inflammatory diseases operates through multiple genes interacting in functional pathways.     

Genetic risk score predicting risk of rheumatoid arthritis phenotypes and age of symptom onset

Background

Cumulative genetic profiles can help identify individuals at high-risk for developing RA. We examined the impact of 39 validated genetic risk alleles on the risk of RA phenotypes characterized by serologic and erosive status.

Methods/Principal Findings

We evaluated single nucleotide polymorphisms at 31 validated RA risk loci and 8 Human Leukocyte Antigen alleles among 542 Caucasian RA cases and 551 Caucasian controls from Nurses'' Health Study and Nurses'' Health Study II. We created a weighted genetic risk score (GRS) and evaluated it as 7 ordinal groups using logistic regression (adjusting for age and smoking) to assess the relationship between GRS group and odds of developing seronegative (RF− and CCP−), seropositive (RF+ or CCP+), erosive, and seropositive, erosive RA phenotypes. In separate case only analyses, we assessed the relationships between GRS and age of symptom onset.In 542 RA cases, 317 (58%) were seropositive, 163 (30%) had erosions and 105 (19%) were seropositive with erosions. Comparing the highest GRS risk group to the median group, we found an OR of 1.2 (95% CI = 0.8–2.1) for seronegative RA, 3.0 (95% CI = 1.9–4.7) for seropositive RA, 3.2 (95% CI = 1.8–5.6) for erosive RA, and 7.6 (95% CI = 3.6–16.3) for seropositive, erosive RA. No significant relationship was seen between GRS and age of onset.

Conclusions/Significance

Results suggest that seronegative and seropositive/erosive RA have different genetic architecture and support the importance of considering RA phenotypes in RA genetic studies.     

Immunochip Identifies Novel,and Replicates Known,Genetic Risk Loci for Rheumatoid Arthritis in Black South Africans

The aim of this study was to identify genetic variants associated with rheumatoid arthritis (RA) risk in black South Africans. Black South African RA patients (n = 263) were compared with healthy controls (n = 374). Genotyping was performed using the Immunochip, and four-digit high-resolution human leukocyte antigen (HLA) typing was performed by DNA sequencing of exon 2. Standard quality control measures were implemented on the data. The strongest associations were in the intergenic region between the HLA-DRB1 and HLA-DQA1 loci. After conditioning on HLA-DRB1 alleles, the effect in the rest of the extended major histocompatibility (MHC) diminished. Non-HLA single nucleotide polymorphisms (SNPs) in the intergenic regions LOC389203|RBPJ, LOC100131131|IL1R1, KIAA1919|REV3L, LOC643749|TRAF3IP2, and SNPs in the intron and untranslated regions (UTR) of IRF1 and the intronic region of ICOS and KIAA1542 showed association with RA (p < 5 × 10⁻⁵). Of the SNPs previously associated with RA in Caucasians, one SNP, rs874040, locating to the intergenic region LOC389203|RBPJ was replicated in this study. None of the variants in the PTPN22 gene was significantly associated. The seropositive subgroups showed similar results to the overall cohort. The effects observed across the HLA region are most likely due to HLA-DRB1, and secondary effects in the extended MHC cannot be detected. Seven non-HLA loci are associated with RA in black South Africans. Similar to Caucasians, the intergenic region between LOC38920 and RBPJ is associated with RA in this population. The strong association of the R620W variant of the PTPN22 gene with RA in Caucasians was not replicated since this variant was monomorphic in our study, but other SNP variants of the PTPN22 gene were also not associated with RA in black South Africans, suggesting that this locus does not play a major role in RA in this population.     

A Robust Genome‐Wide Scan Statistic of the Wellcome Trust Case–Control Consortium

Summary In genome‐wide association (GWA) studies, test statistics that are efficient and robust across various genetic models are preferable, particularly for studying multiple diseases in the Wellcome Trust Case–Control Consortium ( WTCCC, 2007 , Nature 447 , 661–678). A new test statistic, the minimum of the p‐values of the trend test and Pearson's test, was considered by the WTCCC. It is referred to here as MIN2. Because the minimum of two p‐values is no longer a valid p‐value itself, the WTCCC only used it to rank single nucleotide polymorphisms (SNPs) but did not report the p‐values of the associated SNPs when MIN2 was used for ranking. Given its importance in practice, we derive the asymptotic null distribution of MIN2, study some of its analytical properties related to GWA studies, and compare it with existing methods (the trend test, Pearson's test, MAX3, and the constrained likelihood ratio test [CLRT]) by simulations across a wide range of possible genetic models: the recessive (REC), additive (ADD), multiplicative (MUL), dominant (DOM), and overdominant models. The results show that MAX3 and CLRT have greater efficiency robustness than other tests when the REC, ADD/MUL, and DOM models are possible, whereas Pearson's test and MIN2 have greater efficiency robustness if the possible genetic models also include the overdominant model. We conclude that robust tests (MAX3, MIN2, CLRT, and Pearson's test) are preferable to a single trend test for initial GWA studies. The four robust tests are applied to more than 100 SNPs associated with 11 common diseases identified by the two WTCCC GWA studies.     

A Comprehensive Genetic Approach for Improving Prediction of Skin Cancer Risk in Humans

Prediction of genetic risk for disease is needed for preventive and personalized medicine. Genome-wide association studies have found unprecedented numbers of variants associated with complex human traits and diseases. However, these variants explain only a small proportion of genetic risk. Mounting evidence suggests that many traits, relevant to public health, are affected by large numbers of small-effect genes and that prediction of genetic risk to those traits and diseases could be improved by incorporating large numbers of markers into whole-genome prediction (WGP) models. We developed a WGP model incorporating thousands of markers for prediction of skin cancer risk in humans. We also considered other ways of incorporating genetic information into prediction models, such as family history or ancestry (using principal components, PCs, of informative markers). Prediction accuracy was evaluated using the area under the receiver operating characteristic curve (AUC) estimated in a cross-validation. Incorporation of genetic information (i.e., familial relationships, PCs, or WGP) yielded a significant increase in prediction accuracy: from an AUC of 0.53 for a baseline model that accounted for nongenetic covariates to AUCs of 0.58 (pedigree), 0.62 (PCs), and 0.64 (WGP). In summary, prediction of skin cancer risk could be improved by considering genetic information and using a large number of single-nucleotide polymorphisms (SNPs) in a WGP model, which allows for the detection of patterns of genetic risk that are above and beyond those that can be captured using family history. We discuss avenues for improving prediction accuracy and speculate on the possible use of WGP to prospectively identify individuals at high risk.     

Characterizing associations and SNP-environment interactions for GWAS-identified prostate cancer risk markers--results from BPC3

Genome-wide association studies (GWAS) have identified multiple single nucleotide polymorphisms (SNPs) associated with prostate cancer risk. However, whether these associations can be consistently replicated, vary with disease aggressiveness (tumor stage and grade) and/or interact with non-genetic potential risk factors or other SNPs is unknown. We therefore genotyped 39 SNPs from regions identified by several prostate cancer GWAS in 10,501 prostate cancer cases and 10,831 controls from the NCI Breast and Prostate Cancer Cohort Consortium (BPC3). We replicated 36 out of 39 SNPs (P-values ranging from 0.01 to 10⁻²⁸). Two SNPs located near KLK3 associated with PSA levels showed differential association with Gleason grade (rs2735839, P = 0.0001 and rs266849, P = 0.0004; case-only test), where the alleles associated with decreasing PSA levels were inversely associated with low-grade (as defined by Gleason grade <8) tumors but positively associated with high-grade tumors. No other SNP showed differential associations according to disease stage or grade. We observed no effect modification by SNP for association with age at diagnosis, family history of prostate cancer, diabetes, BMI, height, smoking or alcohol intake. Moreover, we found no evidence of pair-wise SNP-SNP interactions. While these SNPs represent new independent risk factors for prostate cancer, we saw little evidence for effect modification by other SNPs or by the environmental factors examined.     

Interpretation of Genetic Association Studies: Markers with Replicated Highly Significant Odds Ratios May Be Poor Classifiers

Recent successful discoveries of potentially causal single nucleotide polymorphisms (SNPs) for complex diseases hold great promise, and commercialization of genomics in personalized medicine has already begun. The hope is that genetic testing will benefit patients and their families, and encourage positive lifestyle changes and guide clinical decisions. However, for many complex diseases, it is arguable whether the era of genomics in personalized medicine is here yet. We focus on the clinical validity of genetic testing with an emphasis on two popular statistical methods for evaluating markers. The two methods, logistic regression and receiver operating characteristic (ROC) curve analysis, are applied to our age-related macular degeneration dataset. By using an additive model of the CFH, LOC387715, and C2 variants, the odds ratios are 2.9, 3.4, and 0.4, with p-values of 10⁻¹³, 10⁻¹³, and 10⁻³, respectively. The area under the ROC curve (AUC) is 0.79, but assuming prevalences of 15%, 5.5%, and 1.5% (which are realistic for age groups 80 y, 65 y, and 40 y and older, respectively), only 30%, 12%, and 3% of the group classified as high risk are cases. Additionally, we present examples for four other diseases for which strongly associated variants have been discovered. In type 2 diabetes, our classification model of 12 SNPs has an AUC of only 0.64, and two SNPs achieve an AUC of only 0.56 for prostate cancer. Nine SNPs were not sufficient to improve the discrimination power over that of nongenetic predictors for risk of cardiovascular events. Finally, in Crohn''s disease, a model of five SNPs, one with a quite low odds ratio of 0.26, has an AUC of only 0.66. Our analyses and examples show that strong association, although very valuable for establishing etiological hypotheses, does not guarantee effective discrimination between cases and controls. The scientific community should be cautious to avoid overstating the value of association findings in terms of personalized medicine before their time.     

A polymorphism in <Emphasis Type="Italic">PTPN2</Emphasis> gene is associated with an earlier onset of type 1 diabetes

The Wellcome Trust Case Control Consortium (WTCCC) genome-wide study found association of PTPN2 with three autoimmune diseases, among them is type 1 diabetes (T1D). This result was confirmed by a follow-up study that pointed to new independent signals within the region. However, both studies were performed in patients with an early-onset T1D. We aimed at replicating the previous results and studying the influence of these polymorphisms in the age at T1D debut. We genotyped 439 T1D Spanish subjects (age at onset, 1 to 65 years) and 861 controls for two PTPN2 single nucleotide polymorphisms (SNPs), rs2542151 and rs478582, and studied the effect of both polymorphisms in age at onset through stratified and continuous analyses. The frequency of rs2542151*G carriers was significantly higher in the early-onset group compared with late-onset patients (p = 0.023) and with controls (OR = 1.61 [1.14–2.26]; p = 0.005). No significant differences were found between controls and late-onset patients. The log-rank chi-square test for the Kaplan–Meier plots (carriers of susceptibility allele vs non carriers) was statistically significant (χ _1df² = 4.485; p = 0.034), yielding an earlier disease debut for G carriers. The analysis of the SNP rs478582 did not reach statistical significance. In summary, we replicate the association detected by the WTCCC and propose that the rs2542151*G allele confers risk to an earlier onset of T1D.