Search

BackgroundAs global populations age and become susceptible to neurodegenerative illnesses, new therapies for Alzheimer disease (AD) are urgently needed. Existing data resources for drug discovery and repurposing fail to capture relationships central to the disease’s etiology and response to drugs. ObjectiveWe designed the Alzheimer’s Knowledge Base (AlzKB) to alleviate this need by providing a comprehensive knowledge representation of AD etiology and candidate therapeutics. MethodsWe designed the AlzKB as a large, heterogeneous graph knowledge base assembled using 22 diverse external data sources describing biological and pharmaceutical entities at different levels of organization (eg, chemicals, genes, anatomy, and diseases). AlzKB uses a Web Ontology Language 2 ontology to enforce semantic consistency and allow for ontological inference. We provide a public version of AlzKB and allow users to run and modify local versions of the knowledge base. ResultsAlzKB is freely available on the web and currently contains 118,902 entities with 1,309,527 relationships between those entities. To demonstrate its value, we used graph data science and machine learning to (1) propose new therapeutic targets based on similarities of AD to Parkinson disease and (2) repurpose existing drugs that may treat AD. For each use case, AlzKB recovers known therapeutic associations while proposing biologically plausible new ones. ConclusionsAlzKB is a new, publicly available knowledge resource that enables researchers to discover complex translational associations for AD drug discovery. Through 2 use cases, we show that it is a valuable tool for proposing novel therapeutic hypotheses based on public biomedical knowledge.

Information is the cornerstone of research, from experimental (meta)data and computational processes to complex inventories of reagents and equipment. These 10 simple rules discuss best practices for leveraging laboratory information management systems to transform this large information load into useful scientific findings.

Assumptions are made about the genetic model of single nucleotide polymorphisms (SNPs) when choosing a traditional genetic encoding: additive, dominant, and recessive. Furthermore, SNPs across the genome are unlikely to demonstrate identical genetic models. However, running SNP-SNP interaction analyses with every combination of encodings raises the multiple testing burden. Here, we present a novel and flexible encoding for genetic interactions, the elastic data-driven genetic encoding (EDGE), in which SNPs are assigned a heterozygous value based on the genetic model they demonstrate in a dataset prior to interaction testing. We assessed the power of EDGE to detect genetic interactions using 29 combinations of simulated genetic models and found it outperformed the traditional encoding methods across 10%, 30%, and 50% minor allele frequencies (MAFs). Further, EDGE maintained a low false-positive rate, while additive and dominant encodings demonstrated inflation. We evaluated EDGE and the traditional encodings with genetic data from the Electronic Medical Records and Genomics (eMERGE) Network for five phenotypes: age-related macular degeneration (AMD), age-related cataract, glaucoma, type 2 diabetes (T2D), and resistant hypertension. A multi-encoding genome-wide association study (GWAS) for each phenotype was performed using the traditional encodings, and the top results of the multi-encoding GWAS were considered for SNP-SNP interaction using the traditional encodings and EDGE. EDGE identified a novel SNP-SNP interaction for age-related cataract that no other method identified: rs7787286 (MAF: 0.041; intergenic region of chromosome 7)-rs4695885 (MAF: 0.34; intergenic region of chromosome 4) with a Bonferroni LRT p of 0.018. A SNP-SNP interaction was found in data from the UK Biobank within 25 kb of these SNPs using the recessive encoding: rs60374751 (MAF: 0.030) and rs6843594 (MAF: 0.34) (Bonferroni LRT p: 0.026). We recommend using EDGE to flexibly detect interactions between SNPs exhibiting diverse action.

Papers describing software are an important part of computational fields of scientific research. These "software papers" are unique in a number of ways, and they require special consideration to improve their impact on the scientific community and their efficacy at conveying important information. Here, we discuss 10 specific rules for writing software papers, covering some of the different scenarios and publication types that might be encountered, and important questions from which all computational researchers would benefit by asking along the way.

Increasingly complex statistical models are being used for the analysis of biological data. Recent commentary has focused on the ability to compute the same outcome for a given dataset (reproducibility). We argue that a reproducible statistical analysis is not necessarily valid because of unique patterns of nonindependence in every biological dataset. We advocate that analyses should be evaluated with known-truth simulations that capture biological reality, a process we call "analysis validation." We review the process of validation and suggest criteria that a validation project should meet. We find that different fields of science have historically failed to meet all criteria, and we suggest ways to implement meaningful validation in training and practice.

Cloud computing has revolutionized the development and operations of hardware and software across diverse technological arenas, yet academic biomedical research has lagged behind despite the numerous and weighty advantages that cloud computing offers. Biomedical researchers who embrace cloud computing can reap rewards in cost reduction, decreased development and maintenance workload, increased reproducibility, ease of sharing data and software, enhanced security, horizontal and vertical scalability, high availability, a thriving technology partner ecosystem, and much more. Despite these advantages that cloud-based workflows offer, the majority of scientific software developed in academia does not utilize cloud computing and must be migrated to the cloud by the user. In this article, we present 11 quick tips for architecting biomedical informatics workflows on compute clouds, distilling knowledge gained from experience developing, operating, maintaining, and distributing software and virtualized appliances on the world's largest cloud. Researchers who follow these tips stand to benefit immediately by migrating their workflows to cloud computing and embracing the paradigm of abstraction.

One in three people has been infected with Mycobacterium tuberculosis (MTB), and the risk for MTB infection in HIV-infected individuals is even higher. We hypothesized that HIV-positive individuals living in tuberculosis-endemic regions who do not get infected by Mycobacterium tuberculosis are genetically resistant. Using an "experiment of nature" design that proved successful in our previous work, we performed a genome-wide association study of tuberculin skin test positivity using 469 HIV-positive patients from prospective study cohorts of tuberculosis from Tanzania and Uganda to identify genetic loci associated with MTB infection in the context of HIV-infection. Among these individuals, 244 tested were tuberculin skin test (TST) positive either at enrollment or during the >8 year follow up, while 225 were not. We identified a genome-wide significant association between a dominant model of rs877356 and binary TST status in the combined cohort (Odds ratio = 0.2671, p = 1.22x10-8). Association was replicated with similar significance when examining TST induration as a continuous trait. The variant lies in the 5q31.1 region, 57kb downstream from IL9. Two-locus analyses of association of variants near rs877356 showed a haplotype comprised of rs877356 and an IL9 missense variant, rs2069885, had the most significant association (p = 1.59x10-12). We also replicated previously linked loci on chromosomes 2, 5, and 11. IL9 is a cytokine produced by mast cells and TH2 cells during inflammatory responses, providing a possible link between airway inflammation and protection from MTB infection. Our results indicate that studying uninfected, HIV-positive participants with extensive exposure increases the power to detect associations in complex infectious disease.

Populations in sub-Saharan Africa are shifting from rural to increasingly urban. Although the burden of cardiovascular disease is expected to increase with this changing landscape, few large studies have assessed a wide range of risk factors in urban and rural populations, particularly in West Africa. We conducted a cross-sectional, population-based survey of 3317 participants from Ghana (≥18 years old), of whom 2265 (57% female) were from a mid-sized city (Sunyani, population ~250,000) and 1052 (55% female) were from surrounding villages (populations

As the most common type of dementia, Alzheimer’s disease (AD) is a neurodegenerative disorder initially manifested by impaired memory performances. While the diagnosis information indicates a dichotomous status of a patient, memory scores have the potential to capture the continuous nature of the disease progression and may provide more insights into the underlying mechanism. In this work, we performed a targeted genetic study of memory scores on an AD cohort to identify the associations between a set of genes highly expressed in the hippocampal region and seven cognitive scores related to episodic memory. Both main effects and interaction effects of the targeted genetic markers on these correlated memory scores were examined. In addition to well-known AD genetic markers APOE and TOMM40, our analysis identified a new risk gene NAV2 through the gene-level main effect analysis. NAV2 was found to be significantly and consistently associated with all seven episodic memory scores. Genetic interaction analysis also yielded a few promising hits warranting further investigation, especially for the RAVLT list B Score.

Plasminogen activator inhibitor 1 (PAI-1), a major modulator of the fibrinolytic system, is an important factor in cardiovascular disease (CVD) susceptibility and severity. PAI-1 is highly heritable, but the few genes associated with it explain only a small portion of its variation. Studies of PAI-1 typically employ linear regression to estimate the effects of genetic variants on PAI-1 levels, but PAI-1 is not normally distributed, even after transformation. Therefore, alternative statistical methods may provide greater power to identify important genetic variants. Additionally, most genetic studies of PAI-1 have been performed on populations of European descent, limiting the generalizability of their results. We analyzed >30,000 variants for association with PAI-1 in a Ghanaian population, using median regression, a non-parametric alternative to linear regression. Three variants associated with median PAI-1, the most significant of which was in the gene arylsulfatase B (ARSB) (p = 1.09 x 10(-7)). We also analyzed the upper quartile of PAI-1, the most clinically relevant part of the distribution, and found 19 SNPs significantly associated in this quartile. Of note an association was found in period circadian clock 3 (PER3). Our results reveal novel associations with median and elevated PAI-1 in an understudied population. The lack of overlap between the two analyses indicates that the genetic effects on PAI-1 are not uniform across its distribution. They also provide evidence of the generalizability of the circadian pathway's effect on PAI-1, as a recent meta-analysis performed in Caucasian populations identified another circadian clock gene (ARNTL).

It is challenging to cluster cancer patients of a certain histopathological type into molecular subtypes of clinical importance and identify gene signatures directly relevant to the subtypes. Current clustering approaches have inherent limitations, which prevent them from gauging the subtle heterogeneity of the molecular subtypes. In this paper we present a new framework: SPARCoC (Sparse-CoClust), which is based on a novel Common-background and Sparse-foreground Decomposition (CSD) model and the Maximum Block Improvement (MBI) co-clustering technique. SPARCoC has clear advantages compared with widely-used alternative approaches: hierarchical clustering (Hclust) and nonnegative matrix factorization (NMF). We apply SPARCoC to the study of lung adenocarcinoma (ADCA), an extremely heterogeneous histological type, and a significant challenge for molecular subtyping. For testing and verification, we use high quality gene expression profiling data of lung ADCA patients, and identify prognostic gene signatures which could cluster patients into subgroups that are significantly different in their overall survival (with p-values < 0.05). Our results are only based on gene expression profiling data analysis, without incorporating any other feature selection or clinical information; we are able to replicate our findings with completely independent datasets. SPARCoC is broadly applicable to large-scale genomic data to empower pattern discovery and cancer gene identification.

Many developmental, physiological, and behavioral processes depend on the precise expression of genes in space and time. Such spatiotemporal gene expression phenotypes arise from the binding of sequence-specific transcription factors (TFs) to DNA, and from the regulation of nearby genes that such binding causes. These nearby genes may themselves encode TFs, giving rise to a transcription factor network (TFN), wherein nodes represent TFs and directed edges denote regulatory interactions between TFs. Computational studies have linked several topological properties of TFNs - such as their degree distribution - with the robustness of a TFN's gene expression phenotype to genetic and environmental perturbation. Another important topological property is assortativity, which measures the tendency of nodes with similar numbers of edges to connect. In directed networks, assortativity comprises four distinct components that collectively form an assortativity signature. We know very little about how a TFN's assortativity signature affects the robustness of its gene expression phenotype to perturbation. While recent theoretical results suggest that increasing one specific component of a TFN's assortativity signature leads to increased phenotypic robustness, the biological context of this finding is currently limited because the assortativity signatures of real-world TFNs have not been characterized. It is therefore unclear whether these earlier theoretical findings are biologically relevant. Moreover, it is not known how the other three components of the assortativity signature contribute to the phenotypic robustness of TFNs. Here, we use publicly available DNaseI-seq data to measure the assortativity signatures of genome-wide TFNs in 41 distinct human cell and tissue types. We find that all TFNs share a common assortativity signature and that this signature confers phenotypic robustness to model TFNs. Lastly, we determine the extent to which each of the four components of the assortativity signature contributes to this robustness.

BACKGROUND:A growing body of evidence suggests that exposure to toxic metals occurs through diet but few studies have comprehensively examined dietary sources of exposure in US populations. PURPOSE:Our goal was to perform a novel dietary-wide association study (DWAS) to identify specific dietary sources of lead, cadmium, mercury, and arsenic exposure in US children and adults. METHODS:We combined data from the National Health and Nutrition Examination Survey with data from the US Department of Agriculture's Food Intakes Converted to Retail Commodities Database to examine associations between 49 different foods and environmental metal exposure. Using blood and urinary biomarkers for lead, cadmium, mercury, and arsenic, we compared sources of dietary exposure among children to that of adults. RESULTS:Diet accounted for more of the variation in mercury and arsenic than lead and cadmium. For instance we estimate 4.5% of the variation of mercury among children and 10.5% among adults is explained by diet. We identified a previously unrecognized association between rice consumption and mercury in a US study population--adjusted for other dietary sources such as seafood, an increase of 10 g/day of rice consumption was associated with a 4.8% (95% CI: 3.6, 5.2) increase in blood mercury concentration. Associations between diet and metal exposure were similar among children and adults, and we recapitulated other known dietary sources of exposure. CONCLUSION:Utilizing this combination of data sources, this approach has the potential to identify and monitor dietary sources of metal exposure in the US population.

Mitochondrial DNA (mtDNA) variation can affect phenotypic variation; therefore, knowing its distribution within and among individuals is of importance to understanding many human diseases. Intra-individual mtDNA variation (heteroplasmy) has been generally assumed to be random. We used massively parallel sequencing to assess heteroplasmy across ten tissues and demonstrate that in unrelated individuals there are tissue-specific, recurrent mutations. Certain tissues, notably kidney, liver and skeletal muscle, displayed the identical recurrent mutations that were undetectable in other tissues in the same individuals. Using RFLP analyses we validated one of the tissue-specific mutations in the two sequenced individuals and replicated the patterns in two additional individuals. These recurrent mutations all occur within or in very close proximity to sites that regulate mtDNA replication, strongly implying that these variations alter the replication dynamics of the mutated mtDNA genome. These recurrent variants are all independent of each other and do not occur in the mtDNA coding regions. The most parsimonious explanation of the data is that these frequently repeated mutations experience tissue-specific positive selection, probably through replication advantage.

We present an extension of the two-class multifactor dimensionality reduction (MDR) algorithm that enables detection and characterization of epistatic SNP-SNP interactions in the context of a quantitative trait. The proposed Quantitative MDR (QMDR) method handles continuous data by modifying MDR's constructive induction algorithm to use a T-test. QMDR replaces the balanced accuracy metric with a T-test statistic as the score to determine the best interaction model. We used a simulation to identify the empirical distribution of QMDR's testing score. We then applied QMDR to genetic data from the ongoing prospective Prevention of Renal and Vascular End-Stage Disease (PREVEND) study.

Systemic lupus erythematosus (SLE) is an inflammatory autoimmune disease with a strong genetic component. African-Americans (AA) are at increased risk of SLE, but the genetic basis of this risk is largely unknown. To identify causal variants in SLE loci in AA, we performed admixture mapping followed by fine mapping in AA and European-Americans (EA). Through genome-wide admixture mapping in AA, we identified a strong SLE susceptibility locus at 2q22-24 (LOD=6.28), and the admixture signal is associated with the European ancestry (ancestry risk ratio ~1.5). Large-scale genotypic analysis on 19,726 individuals of African and European ancestry revealed three independently associated variants in the IFIH1 gene: an intronic variant, rs13023380 [P(meta) = 5.20×10(-14); odds ratio, 95% confidence interval = 0.82 (0.78-0.87)], and two missense variants, rs1990760 (Ala946Thr) [P(meta) = 3.08×10(-7); 0.88 (0.84-0.93)] and rs10930046 (Arg460His) [P(dom) = 1.16×10(-8); 0.70 (0.62-0.79)]. Both missense variants produced dramatic phenotypic changes in apoptosis and inflammation-related gene expression. We experimentally validated function of the intronic SNP by DNA electrophoresis, protein identification, and in vitro protein binding assays. DNA carrying the intronic risk allele rs13023380 showed reduced binding efficiency to a cellular protein complex including nucleolin and lupus autoantigen Ku70/80, and showed reduced transcriptional activity in vivo. Thus, in SLE patients, genetic susceptibility could create a biochemical imbalance that dysregulates nucleolin, Ku70/80, or other nucleic acid regulatory proteins. This could promote antibody hypermutation and auto-antibody generation, further destabilizing the cellular network. Together with molecular modeling, our results establish a distinct role for IFIH1 in apoptosis, inflammation, and autoantibody production, and explain the molecular basis of these three risk alleles for SLE pathogenesis.

Genome-wide association studies (GWAS) have evolved over the last ten years into a powerful tool for investigating the genetic architecture of human disease. In this work, we review the key concepts underlying GWAS, including the architecture of common diseases, the structure of common human genetic variation, technologies for capturing genetic information, study designs, and the statistical methods used for data analysis. We also look forward to the future beyond GWAS.

Bladder cancer is the 4(th) most common cancer among men in the U.S. We analyzed variant genotypes hypothesized to modify major biological processes involved in bladder carcinogenesis, including hormone regulation, apoptosis, DNA repair, immune surveillance, metabolism, proliferation, and telomere maintenance. Logistic regression was used to assess the relationship between genetic variation affecting these processes and susceptibility in 563 genotyped urothelial cell carcinoma cases and 863 controls enrolled in a case-control study of incident bladder cancer conducted in New Hampshire, U.S. We evaluated gene-gene interactions using Multifactor Dimensionality Reduction (MDR) and Statistical Epistasis Network analysis. The 3'UTR flanking variant form of the hormone regulation gene HSD3B2 was associated with increased bladder cancer risk in the New Hampshire population (adjusted OR 1.85 95%CI 1.31-2.62). This finding was successfully replicated in the Texas Bladder Cancer Study with 957 controls, 497 cases (adjusted OR 3.66 95%CI 1.06-12.63). The effect of this prevalent SNP was stronger among males (OR 2.13 95%CI 1.40-3.25) than females (OR 1.56 95%CI 0.83-2.95), (SNP-gender interaction P = 0.048). We also identified a SNP-SNP interaction between T-cell activation related genes GATA3 and CD81 (interaction P = 0.0003). The fact that bladder cancer incidence is 3-4 times higher in males suggests the involvement of hormone levels. This biologic process-based analysis suggests candidate susceptibility markers and supports the theory that disrupted hormone regulation plays a role in bladder carcinogenesis.

The proteins, tissue plasminogen activator (t-PA) and plasminogen activator inhibitor 1 (PAI-1), act in concert to balance thrombus formation and degradation, thereby modulating the development of arterial thrombosis and excessive bleeding. PAI-1 is upregulated by the renin-angiotensin system (RAS), specifically by angiotensin II, the product of angiotensin converting enzyme (ACE) cleavage of angiotensin I, which is produced by the cleavage of angiotensinogen (AGT) by renin (REN). ACE indirectly stimulates the release of t-PA which, in turn, activates the corresponding fibrinolytic system. Single polymorphisms in these pathways have been shown to significantly impact plasma levels of t-PA and PAI-1 differently in Ghanaian males and females. Here we explore the involvement of epistatic interactions between the same polymorphisms in central genes of the RAS and fibrinolytic systems on plasma t-PA and PAI-1 levels within the same population (n = 992). Statistical modeling of pairwise interactions was done using two-way ANOVA between polymorphisms in the ETNK2, RENIN, ACE, PAI-1, t-PA, and AGT genes. The most significant interactions that associated with t-PA levels were between the ETNK2 A6135G and the REN T9435C polymorphisms in females (p = 0.006) and the REN T9435C and the TPA I/D polymorphisms (p = 0.005) in males. The most significant interactions for PAI-1 levels were with REN T9435C and the TPA I/D polymorphisms (p = 0.001) in females, and the association of REN G6567T with the TPA I/D polymorphisms (p = 0.032) in males. Our results provide evidence for multiple genetic effects that may not be detected using single SNP analysis. Because t-PA and PAI-1 have been implicated in cardiovascular disease these results support the idea that the genetic architecture of cardiovascular disease is complex. Therefore, it is necessary to consider the relationship between interacting polymorphisms of pathway specific genes that predict t-PA and PAI-1 levels.

Gene-on-gene regulations are key components of every living organism. Dynamical abstract models of genetic regulatory networks help explain the genome's evolvability and robustness. These properties can be attributed to the structural topology of the graph formed by genes, as vertices, and regulatory interactions, as edges. Moreover, the actual gene interaction of each gene is believed to play a key role in the stability of the structure. With advances in biology, some effort was deployed to develop update functions in boolean models that include recent knowledge. We combine real-life gene interaction networks with novel update functions in a boolean model. We use two sub-networks of biological organisms, the yeast cell-cycle and the mouse embryonic stem cell, as topological support for our system. On these structures, we substitute the original random update functions by a novel threshold-based dynamic function in which the promoting and repressing effect of each interaction is considered. We use a third real-life regulatory network, along with its inferred boolean update functions to validate the proposed update function. Results of this validation hint to increased biological plausibility of the threshold-based function. To investigate the dynamical behavior of this new model, we visualized the phase transition between order and chaos into the critical regime using Derrida plots. We complement the qualitative nature of Derrida plots with an alternative measure, the criticality distance, that also allows to discriminate between regimes in a quantitative way. Simulation on both real-life genetic regulatory networks show that there exists a set of parameters that allows the systems to operate in the critical region. This new model includes experimentally derived biological information and recent discoveries, which makes it potentially useful to guide experimental research. The update function confers additional realism to the model, while reducing the complexity and solution space, thus making it easier to investigate.

Vascular fibrinolytic balance is maintained primarily by interplay of tissue plasminogen activator (t-PA) and plasminogen activator inhibitor type 1 (PAI-1). Previous research has shown that polymorphisms in genes from the renin-angiotensin (RA), bradykinin, and fibrinolytic systems affect plasma concentrations of both t-PA and PAI-1 through a set of gene-gene interactions. In the present study, we extend this finding by exploring the effects of polymorphisms in genes from these systems on incident cardiovascular disease, explicitly examining two-way interactions in a large population-based study.Data from the population-based PREVEND study in Groningen, The Netherlands (n = 8,138) were analyzed. The effects of the polymorphisms and their interactions on cardiovascular events were analyzed via Cox proportional hazards models. There was no association between five of the six polymorphisms singly and risk of cardiovascular disease. There was a significant main effect for the ACE I/D polymorphism for both dominant and additive coding schemes. There were significant interactions between the following polymorphism pairs even after adjustment for known risk factors: ACE I/D & PAI-1 4G/5G (p = 0.012), BDKRB2 C181T & ACE I/D (p = 0.016), BDKRB2 C58T & ACE I/D (p = 0.025), BDKRB2 exon 1 I/D & AT1R A1166C (p = 0.017), and BDKRB2 C58T & AT1R A1166C (p = 0.015).This study suggests possible interactions between genes from the RA, bradykinin, and fibrinolytic systems on the risk of cardiovascular disease, extending previous research that has demonstrated that interactions among genes from these systems influence plasma concentrations of both t-PA and PAI-1. Further explorations of these interactions are needed.

Genetic structure due to ancestry has been well documented among many divergent human populations. However, the ability to associate ancestry with genetic substructure without using supervised clustering has not been explored in more presumably homogeneous and admixed US populations. The goal of this study was to determine if genetic structure could be detected in a United States population from a single state where the individuals have mixed European ancestry. Using Bayesian clustering with a set of 960 single nucleotide polymorphisms (SNPs) we found evidence of population stratification in 864 individuals from New Hampshire that can be used to differentiate the population into six distinct genetic subgroups. We then correlated self-reported ancestry of the individuals with the Bayesian clustering results. Finnish and Russian/Polish/Lithuanian ancestries were most notably found to be associated with genetic substructure. The ancestral results were further explained and substantiated using New Hampshire census data from 1870 to 1930 when the largest waves of European immigrants came to the area. We also discerned distinct patterns of linkage disequilibrium (LD) between the genetic groups in the growth hormone receptor gene (GHR). To our knowledge, this is the first time such an investigation has uncovered a strong link between genetic structure and ancestry in what would otherwise be considered a homogenous US population.

Replication has become the gold standard for assessing statistical results from genome-wide association studies. Unfortunately this replication requirement may cause real genetic effects to be missed. A real result can fail to replicate for numerous reasons including inadequate sample size or variability in phenotype definitions across independent samples. In genome-wide association studies the allele frequencies of polymorphisms may differ due to sampling error or population differences. We hypothesize that some statistically significant independent genetic effects may fail to replicate in an independent dataset when allele frequencies differ and the functional polymorphism interacts with one or more other functional polymorphisms. To test this hypothesis, we designed a simulation study in which case-control status was determined by two interacting polymorphisms with heritabilities ranging from 0.025 to 0.4 with replication sample sizes ranging from 400 to 1600 individuals. We show that the power to replicate the statistically significant independent main effect of one polymorphism can drop dramatically with a change of allele frequency of less than 0.1 at a second interacting polymorphism. We also show that differences in allele frequency can result in a reversal of allelic effects where a protective allele becomes a risk factor in replication studies. These results suggest that failure to replicate an independent genetic effect may provide important clues about the complexity of the underlying genetic architecture. We recommend that polymorphisms that fail to replicate be checked for interactions with other polymorphisms, particularly when samples are collected from groups with distinct ethnic backgrounds or different geographic regions.