Development of GMDR-GPU for gene-gene interaction analysis and its application to WTCCC GWAS data for type 2 diabetes

Review on GPU accelerated methods for genome-wide SNP-SNP interactions

Detecting genome-wide SNP-SNP interactions (epistasis) efficiently is essential to harnessing the vast data now available from modern biobanks. With millions of SNPs and genetic information from hundreds of thousands of individuals, researchers are positioned to uncover new insights into complex disease pathways. However, this data scale brings significant computational and statistical challenges. To address these, recent approaches leverage GPU-based parallel computing for high-throughput, cost-effective analysis and refine algorithms to improve time and memory efficiency. In this survey, we systematically review GPU-accelerated methods for exhaustive epistasis detection, detailing the statistical models used and the computational strategies employed to enhance performance. Our findings indicate substantial speedups with GPU implementations over traditional CPU approaches. We conclude that while GPU-based solutions hold promise for advancing genomic research, continued innovation in both algorithm design and hardware optimization is necessary to meet future data challenges in the field.

Association of common variants of TCF7L2 and PCSK2 with gestational diabetes mellitus in West Bengal, India

The genetic etiology of gestational diabetes mellitus (GDM) was suggested to overlap with type-2 diabetes(T2D). Transcription factor 7-like 2 (TCF7L2) and Proprotein Convertase Subtilisin/Kexin type 2 (PCSK2) are T2D susceptibility genes of the insulin synthesis/processing pathway. We analyzed associations of TCF7L2 and PCSK2 variants with GDM risk and evaluated their potential impact on impaired insulin processing in an eastern Indian population. The study included 114 GDM (case) and 228 non-GDM pregnant women (control). rs7903146, rs4132670, rs12255372 of TCF7L2, and rs2269023 of PCSK2 were genotyped by PCR-RFLP, and genotype distributions were compared between case and control. Fasting serum proinsulin and C-peptide levels were measured by ELISA and the Proinsulin/C-peptide ratio was considered an indicator of proinsulin conversion. Significantly higher frequency of risk allele (T) of rs12255372 (p = 0.02, OR = 2.0, 95%CI = 1.11-3.64) and rs4132670 (p = 0.002, OR = 2.26, 95%CI = 1.32-3.87) of TCF7L2 was found in GDM cases than non-GDM controls; TT genotype was associated with significantly increased disease risk. In rs7903146 (TCF7L2) and rs2269023 (PCSK2), although the frequency of risk allele (T) was not significantly higher in cases than controls, an association of TT for both variants remained significant with higher GDM risk in the recessive model. Increased serum pro-insulin and proinsulin:c-peptide ratio was found in GDM than non-GDM women and the phenomenon showed significant association with careers of risk alleles for TCF7L2 variants. In conclusion, TCF7L2 and PCSK2 variants are related to GDM risk in the studied population and hence may serve as potential biomarkers for assessing the disease risk. TCF7L2 variants contribute to impaired insulin processing.

Genome-wide association studies revealed complex genetic architecture and breeding perspective of maize ear traits

BACKGROUND

Maize (Zea Mays) is one of the world's most important crops. Hybrid maize lines resulted a major improvement in corn production in the previous and current centuries. Understanding the genetic mechanisms of the corn production associated traits greatly facilitate the development of superior hybrid varieties.

RESULT

In this study, four ear traits associated with corn production of Nested Association Mapping (NAM) population were analyzed using a full genetic model, and further, optimal genotype combinations and total genetic effects of current best lines, superior lines, and superior hybrids were predicted for each of the traits at four different locations. The analysis identified 21-34 highly significant SNPs (-log₁₀P > 5), with an estimated total heritability of 37.31-62.34%, while large contributions to variations was due to dominance, dominance-related epistasis, and environmental interaction effects ([Formula: see text] 14.06% ~ 49.28%), indicating these factors contributed significantly to phenotypic variations of the ear traits. Environment-specific genetic effects were also discovered to be crucial for maize ear traits. There were four SNPs found for three ear traits: two for ear length and weight, and two for ear row number and length. Using the Enumeration method and the stepwise tuning technique, optimum multi-locus genotype combinations for superior lines were identified based on the information obtained from GWAS.

CONCLUSIONS

Predictions of genetic breeding values showed that different genotype combinations in different geographical regions may be better, and hybrid-line variety breeding with homozygote and heterozygote genotype combinations may have a greater potential to improve ear traits.

Skeletal muscle RUNX1 is related to insulin sensitivity through its effect on myogenic potential

OBJECTIVE

Skeletal muscle is the major site of insulin action. There are limited data on the relationship between insulin action and skeletal muscle myogenic/regenerative potential. RUNX1 is a transcription factor which plays a role in muscle development and regeneration. The aim of our study was to assess the role of skeletal muscle myogenic/regenerative potential in the development of insulin resistance through the studies on RUNX1 transcription factor.

DESIGN

This study is a cross-sectional study. Experimental part with myoblast cell line culture.

METHODS

We examined 41 young healthy volunteers, 21 normal weight and 20 with overweight or obesity. Hyperinsulinemic-euglycemic clamp and vastus lateralis muscle biopsy were performed. In L6 myoblast and human skeletal muscle myoblasts (hSkMM) cell cultures, RUNX1 was silenced at two stages of development. Cell growth, the expression of markers of myogenesis, nuclei fusion index, Akt phosphorylation and glucose uptake were measured.

RESULTS

Skeletal muscle RUNX1 expression was decreased in overweight/obese individuals in comparison with normal-weight individuals and was positively related to insulin sensitivity, independently of BMI. Runx1 loss-of-function at the stage of myoblast inhibited myoblast proliferation and differentiation and reduced insulin-stimulated Akt phosphorylation and insulin-stimulated glucose uptake. In contrast, Runx1 knockdown in myotubes did not affect Akt phosphorylation, glucose uptake and other parameters studied.

CONCLUSIONS

Myogenic/regenerative potential of adult skeletal muscle may be an important determinant of insulin action. Our data suggest that muscle RUNX1 may play a role in the modulation of insulin action through its effect on myogenesis.

Polymorphisms and Gene-Gene Interaction in AGER/IL6 Pathway Might Be Associated with Diabetic Ischemic Heart Disease

Background: Although the genetic susceptibility to diabetes and ischemic heart disease (IHD) has been well demonstrated, studies aimed at exploring gene variations associated with diabetic IHD are still limited; Methods: Our study included 204 IHD cases who had been diagnosed with diabetes before the diagnosis of IHD and 882 healthy controls. Logistic regression was used to find the association of candidate SNPs and polygenic risk score (PRS) with diabetic IHD. The diagnostic accuracy was represented with AUC. Generalized multifactor dimensionality reduction (GMDR) was used to illustrate gene-gene interactions; Results: For IL6R rs4845625, the CT and TT genotypes were associated with a lower risk of diabetic IHD than the CC genotype (OR = 0.619, p = 0.033; OR = 0.542, p = 0.025, respectively). Haplotypes in the AGER gene (rs184003-rs1035798-rs2070600-rs1800624) and IL6R gene (rs7529229-rs4845625-rs4129267-rs7514452-rs4072391) were both significantly associated with diabetic IHD. PRS was associated with the disease (OR = 1.100, p = 0.005) after adjusting for covariates, and the AUC were 0.763 (p < 0.001). The GMDR analysis suggested that rs184003 and rs4845625 were the best interaction model after permutation testing (p = 0.001) with a cross-validation consistency of 10/10; Conclusions: SNPs and haplotypes in the AGER and IL6R genes and the interaction of rs184003 and rs4845625 were significantly associated with diabetic IHD.

Association analysis between the interaction of RAS family genes mutations and papillary thyroid carcinoma in the Han Chinese population

Papillary thyroid carcinoma (PTC) is the major subtype of thyroid cancer, accounting for 75%-85% of all thyroid malignancies. This study aimed to identify the association between the interactions of single nucleotide polymorphisms (SNPs) in RAS family genes and PTC in the Han Chinese population, to provide clues to the pathogenesis and potential therapeutic targets for PTC. Hap Map and NCBI-db SNP databases were used to retrieve SNPs. Haploview 4.2 software was used to filter SNPs based on specific parameters, six SNPs of RAS gene (KRAS-rs12427141, KRAS-rs712, KRAS-rs7315339, HRAS-rs12628, NRAS-rs14804 and NRAS-rs2273267) were genotyped by matrix-assisted laser desorption/ionization time of flight mass spectrometry (MALDI-TOF-MS) in 673 PTC patients and 657 healthy controls, the interactive effect was evaluated by crossover analysis, logistic regression and GMDR software. We found that genetic mutation in rs712 have significant associations with PTC risk after Bonferroni correction (p<0.001). The interaction between KRAS-rs12427141 and HRAS-rs12628 increased the risk of PTC (U=-2.119, p<0.05), the interaction between KRAS-rs2273267 and HRAS-rs7315339 reduced the risk of PTC (U=2.195, p<0.05). GMDR analysis showed that the two-factor model (KRAS-rs712, NRAS-rs2273267) was the best (p=0.0107). Summarily, there are PTC-related interactions between RAS family genes polymorphisms in the Han Chinese population.

Genome-Wide Search for SNP Interactions in GWAS Data: Algorithm, Feasibility, Replication Using Schizophrenia Datasets

In this study, we looked for potential gene-gene interaction in susceptibility to schizophrenia by an exhaustive searching for SNP-SNP interactions in 3 GWAS datasets (phs000021:phg000013, phs000021:phg000014, phs000167) using our recently published algorithm. The search space for SNP-SNP interaction was confined to 8 biologically plausible ways of interaction under dominant-dominant or recessive-recessive modes. First, we performed our search of all pair-wise combination of 729,454 SNPs after filtering by SNP genotype quality. All possible pairwise interactions of any 2 SNPs (5 × 10¹¹) were exhausted to search for significant interaction which was defined by p-value of chi-square tests. Nine out the top 10 interactions, protein coding genes were partnered with non-coding RNA (ncRNA) which suggested a new alternative insight into interaction biology other than the frequently sought-after protein-protein interaction. Therefore, we extended to look for replication among the top 10,000 interaction SNP pairs and high proportion of concurrent genes forming the interaction pairs were found. The results indicated that an enrichment of signals over noise was present in the top 10,000 interactions. Then, replications of SNP-SNP interaction were confirmed for 14 SNPs-pairs in both replication datasets. Biological insight was highlighted by a potential binding between FHIT (protein coding gene) and LINC00969 (lncRNA) which showed a replicable interaction between their SNPs. Both of them were reported to have expression in brain. Our study represented an early attempt of exhaustive interaction analysis of GWAS data which also yield replicated interaction and new insight into understanding of genetic interaction in schizophrenia.

Robustification of Linear Regression and Its Application in Genome-Wide Association Studies

Regression analysis is one of the most popular statistical techniques that attempt to explore the relationships between a response (dependent) variable and one or more explanatory (independent) variables. To test the overall significance of regression, F-statistic is used if the parameters are estimated by the least-squares estimators (LSEs), while if the parameters are estimated by the maximum likelihood estimators (MLEs), the likelihood ratio test (LRT) statistic is used. However, both procedures produce misleading results and often fail to provide good fits to the reasonable space of the dataset in the presence of outlying observations. Moreover, outliers occur very frequently in any real datasets as well as in the molecular OMICS datasets. Hence, an effort is made in this study to robustify MLE based regression analysis by maximizing the β-likelihood function. The tuning parameter β is selected by cross-validation. For β = 0, the proposed method reduces to the classical MLE based regression analysis. We inspect the performance of the proposed method using both synthetic and real data analysis. The results of simulations indicate that the proposed method performs better than traditional methods in both outliers and high leverage points to estimate the parameters and mean square errors. The results of relative efficiency analysis show that the proposed estimator is relatively less affected than the popular estimators, including S, MM, and fast-S for normal error distribution in case high dimension and outliers. Also, real data analysis results demonstrated that the proposed method shows robust properties with respect to data contaminations, overcome the drawback of the traditional methods. Genome-wide association studies (GWAS) by the proposed method identify the vital gene influencing hypertension and iron level in the liver and spleen of mice. Furthermore, we have identified 15 and 21 significant SNPs for chalkiness degree and chalkiness percentage, respectively, by GWAS based on the proposed method. The variant of the SNPs might be provided the new resources for grain quality traits and could be used for further molecular and physiological analysis to enhance the better quality of rice grain. These results offer an important basis for further understanding of the robust regression analysis, which might be applied in various fields, including business, genetics, and bioinformatics.

Genetic Modifiers and Rare Mendelian Disease

Despite advances in high-throughput sequencing that have revolutionized the discovery of gene defects in rare Mendelian diseases, there are still gaps in translating individual genome variation to observed phenotypic outcomes. While we continue to improve genomics approaches to identify primary disease-causing variants, it is evident that no genetic variant acts alone. In other words, some other variants in the genome (genetic modifiers) may alleviate (suppress) or exacerbate (enhance) the severity of the disease, resulting in the variability of phenotypic outcomes. Thus, to truly understand the disease, we need to consider how the disease-causing variants interact with the rest of the genome in an individual. Here, we review the current state-of-the-field in the identification of genetic modifiers in rare Mendelian diseases and discuss the potential for future approaches that could bridge the existing gap.

Genetic Susceptibility to Gestational Diabetes Mellitus in a Chinese Population

Introduction: New genetic variants associated with susceptibility to obesity and metabolic diseases have been discovered in recent genome-wide association (GWA) studies. The aim of this study was to investigate the association of theses risk variants with gestational diabetes mellitus (GDM). Methods: We performed a case-control study including 964 unrelated pregnant women with GDM and 1,021 pregnant women with normal glucose tolerance (as controls). A total of 33 genetic variants confirmed by GWA studies for obesity and metabolic diseases were selected and measured. Results: We observed that FTO rs1121980 and KCNQ1 rs163182 conferred a decreased GDM risk in the dominant and additive model [additive model: OR (95% CI) = 0.79 (0.67-0.94), P = 0.007 for rs1121980; OR(95%CI) = 0.84 (0.73-0.96), P = 0.009 for rs163182], whereas MC4R rs12970134 and PROX1 rs340841 conferred an increased GDM risk in the dominant, recessive, and additive model [additive model: OR(95%CI) = 1.25 (1.07-1.46), P = 0.006 for rs12970134; OR(95%CI) = 1.22 (1.07-1.39), P = 0.002 for rs340841). With the increasing number of risk alleles of the four significant SNPs, GDM risk was significantly increased in a dose-dependent manner (P_trend < 0.001). And the significant positive associations between the weighted genetic risk score and risk of GDM persisted. Further function annotation indicated that these four SNPs may fall on the functional elements of human pancreatic islets. The genotype-phenotype associations indicated that these SNPs may contribute to GDM by affecting the expression levels of their nearby or distant genes. Conclusion: Our study suggests that FTO rs1121980, KCNQ1 rs163182, MC4R rs12970134, and PROX1 rs340841 may be markers for susceptibility to GDM in a Chinese population.

Single-cell transcriptomics unveils gene regulatory network plasticity

BACKGROUND

Single-cell RNA sequencing (scRNA-seq) plays a pivotal role in our understanding of cellular heterogeneity. Current analytical workflows are driven by categorizing principles that consider cells as individual entities and classify them into complex taxonomies.

RESULTS

We devise a conceptually different computational framework based on a holistic view, where single-cell datasets are used to infer global, large-scale regulatory networks. We develop correlation metrics that are specifically tailored to single-cell data, and then generate, validate, and interpret single-cell-derived regulatory networks from organs and perturbed systems, such as diabetes and Alzheimer's disease. Using tools from graph theory, we compute an unbiased quantification of a gene's biological relevance and accurately pinpoint key players in organ function and drivers of diseases.

CONCLUSIONS

Our approach detects multiple latent regulatory changes that are invisible to single-cell workflows based on clustering or differential expression analysis, significantly broadening the biological insights that can be obtained with this leading technology.

Statistical Methods and Software for Substance Use and Dependence Genetic Research

BACKGROUND

Substantial substance use disorders and related health conditions emerged dur-ing the mid-20th century and continue to represent a remarkable 21st century global burden of disease. This burden is largely driven by the substance-dependence process, which is a complex process and is influenced by both genetic and environmental factors. During the past few decades, a great deal of pro-gress has been made in identifying genetic variants associated with Substance Use and Dependence (SUD) through linkage, candidate gene association, genome-wide association and sequencing studies.

METHODS

Various statistical methods and software have been employed in different types of SUD ge-netic studies, facilitating the identification of new SUD-related variants.

CONCLUSION

In this article, we review statistical methods and software that are currently available for SUD genetic studies, and discuss their strengths and limitations.

Interaction between TP53 and XRCC1 increases susceptibility to cervical cancer development: a case control study

BACKGROUND

Cervical cancer is the 4th highest cause of female reproductive tract malignancies. Multiple loci have been identified as important determinant factors for tumor susceptibility. In this report, we aimed to explore the roles of gene polymorphisms affecting x-ray repair cross complementing 1 (XRCC1), the tumor protein p53 (TP53), and fibroblast growth factor receptor 3 (FGFR3) in the context of susceptibility to cervical cancer. Additionally, we assessed the impact of single nucleotide polymorphism-single nucleotide polymorphism (SNP-SNP) interaction of these three genes in the context of cervical cancer risk in Chinese women.

METHODS

A case-control study consisted of 340 women located in Chongqing. Of these women, 121 were diagnosed with cervical cancer, 118 served as healthy controls, and 101 were specifically recruited elderly patients above the age of 80 who showed no history of cervical cancer. Three SNPs (XRCC1 rs25487, TP53 rs1042522, and FGFR3 rs121913483) were examined using mutation analysis of mismatch amplification PCR (MAMA-PCR) on samples obtained from peripheral blood.

RESULTS

Our results indicated that females from southwestern China all exhibited a wild-type phenotype at FGFR3 rs121913483. We also observed that the rs25487 mutation was significantly increased within the cervical cancer population. A 2-locus SNP-SNP interaction pattern (rs25487 and rs1042522) was significantly associated with cervical cancer risk (cases vs. negative controls: OR = 4.63, 95% CI = 1.83-11.75; cases vs. elderly group: OR = 17.61, 95% CI = 4.34-71.50).

CONCLUSIONS

This is the first study to identify a novel interaction between the XRCC1 and TP53 genes that is highly associated with susceptibility to cervical cancer risk in a female population in southwestern China.

Genome-wide association study of maize plant architecture using F1 populations

Genome-wide association study of maize plant architecture using F₁ populations can better dissect various genetic effects that can provide precise guidance for genetic improvement in maize breeding. Maize grain yield has increased at least eightfold during the past decades. Plant architecture, including plant height, leaf angle, leaf length, and leaf width, has been changed significantly to adapt to higher planting density. Although the genetic architecture of these traits has been dissected using different populations, the genetic basis remains unclear in the F₁ population. In this work, we perform a genome-wide association study of the four traits using 573 F₁ hybrids with a mixed linear model approach and QTXNetwork mapping software. A total of 36 highly significant associated quantitative trait SNPs were identified for these traits, which explained 51.86-79.92% of the phenotypic variation and were contributed mainly by additive, dominance, and environment-specific effects. Heritability as a result of environmental interaction was more important for leaf angle and leaf length, while major effects (a, aa, and d) were more important for leaf width and plant height. The potential breeding values of the superior lines and superior hybrids were also predicted, and these values can be applied in maize breeding by direct selection of superior genotypes for the associated quantitative trait SNPs. A total of 108 candidate genes were identified for the four traits, and further analysis was performed to screen the potential genes involved in the development of maize plant architecture. Our results provide new insights into the genetic architecture of the four traits, and will be helpful in marker-assisted breeding for maize plant architecture.

Association and cis-mQTL analysis of variants in serotonergic genes associated with nicotine dependence in Chinese Han smokers

Variants in serotonergic genes are implicated in nicotine dependence (ND) in subjects of European and African origin, but their involvement with smoking in Asians is largely unknown. Moreover, mechanisms underlying the ND risk-associated single-nucleotide polymorphisms (SNPs) in these genes are rarely investigated. The Fagerström Test for Nicotine Dependence (FTND) score was used to assess ND in 2616 male Chinese Han smokers. Both association and interaction analysis were used to examine the association of variants in the serotonergic genes with FTND. Further, expression and methylation quantitative trait loci (cis-mQTL) analysis was employed to determine the association of individual SNPs with the extent of methylation of each CpG locus. Individual SNP-based association analysis revealed that rs1176744 in HTR3B was marginally associated with FTND (p = 0.042). Haplotype-based association analysis found that one major haplotype, T-T-A-G, formed by SNPs rs3758987-rs4938056-rs1176744-rs2276305, located in the 5' region of HTR3B, showed a significant association with FTND (p = 0.00025). Further, a significant genetic interactive effect affecting ND was detected among SNPs rs10160548 in HTR3A, and rs3758987, rs2276305, and rs1672717 in HTR3B (p = 0.0074). Finally, we found four CpG sites (CpG_4543549, CpG_4543464, CpG_4543682, and CpG_4546888) to be significantly associated with three cis-mQTL SNPs (i.e., rs3758987, rs4938056, and rs1176744) located in our detected haplotype within HTR3B. In sum, we showed SNP rs1176744 (Tyr129Ser) to be associated with ND. Together with the SNPs rs3758987 and rs4938056 in HTR3B, they formed a major haplotype, which had significant association with ND. We further showed these SNPs contribute to ND through four methylated sites in HTR3B. All these findings suggest that variants in the serotonergic system play an important role in ND in the Chinese Han population. More importantly, these findings demonstrated that the involvement of this system in ND is through gene-by-gene interaction and methylation.

Genetic risk score based on fat mass and obesity-associated, transmembrane protein 18 and fibronectin type III domain containing 5 polymorphisms is associated with anthropometric characteristics in South Brazilian children and adolescents

The prevalence of childhood obesity has increased worldwide. Although it is considered a polygenic inheritance disease, little is known about its susceptibility when the additive effect is considered. The aim of this study is to investigate whether the genetic risk score (GRS) based on previously associated obesity polymorphisms (SNP) rs9939609 (fat mass and obesity-associated (FTO)), rs6548238 (transmembrane protein 18 (TMEM18)) and rs16835198 (fibronectin type III domain containing 5 (FNDC5)) could serve as a predictor for anthropometric characteristics in a sample of Brazilian children and adolescents. This is a cross-sectional study with 1471 children and adolescents aged 6–17 years. BMI, waist circumference (WC) and percentage of body fat and metabolic parameters were verified. In all, three SNP were genotyped by TaqMan™ allelic discrimination. The metabolic and anthropometric parameters were compared between the genotypes, and the unweighted and weighted GRS (GRS and wGRS, respectively) were created to test the additive effect of these genetic polymorphisms on anthropometric parameters. The prevalence of overweight plus obesity was 41 %. Significant associations were identified forFTOrs9939609,TMEM18rs6548238 andFNDC5rs16835198 and for GRS and wGRS with anthropometric phenotypes. The higher score of wGRS was associated with obesity (OR: 2·65, 95 % CI 1·40, 5·04,P=0·003) and with greater WC (OR: 2·91, 95 % CI 1·57, 5·40,P=0·001). Our results suggest that these genetic variants contribute to obesity susceptibility in children and adolescents and reinforce the idea that the additive effect may be useful to elucidate the genetic component of obesity.

Discovering Genetic Factors for psoriasis through exhaustively searching for significant second order SNP-SNP interactions

In this paper, we aim at discovering genetic factors of psoriasis through searching for statistically significant SNP-SNP interactions exhaustively from two real psoriasis genome-wide association study datasets (phs000019.v1.p1 and phs000982.v1.p1) downloaded from the database of Genotypes and Phenotypes. To deal with the enormous search space, our search algorithm is accelerated with eight biological plausible interaction patterns and a pre-computed look-up table. After our search, we have discovered several SNPs having a stronger association to psoriasis when they are in combination with another SNP and these combinations may be non-linear interactions. Among the top 20 SNP-SNP interactions being found in terms of pairwise p-value and improvement metric value, we have discovered 27 novel potential psoriasis-associated SNPs where most of them are reported to be eQTLs of a number of known psoriasis-associated genes. On the other hand, we have inferred a gene network after selecting the top 10000 SNP-SNP interactions in terms of improvement metric value and we have discovered a novel long distance interaction between XXbac-BPG154L12.4 and RNU6-283P which is not a long distance haplotype and may be a new discovery. Finally, our experiments with the synthetic datasets have shown that our pre-computed look-up table technique can significantly speed up the search process.

Evolutionary footprint of epistasis

Variation of an inherited trait across a population cannot be explained by additive contributions of relevant genes, due to epigenetic effects and biochemical interactions (epistasis). Detecting epistasis in genomic data still represents a significant challenge that requires a better understanding of epistasis from the mechanistic point of view. Using a standard Wright-Fisher model of bi-allelic asexual population, we study how compensatory epistasis affects the process of adaptation. The main result is a universal relationship between four haplotype frequencies of a single site pair in a genome, which depends only on the epistasis strength of the pair defined regarding Darwinian fitness. We demonstrate the existence, at any time point, of a quasi-equilibrium between epistasis and disorder (entropy) caused by random genetic drift and mutation. We verify the accuracy of these analytic results by Monte-Carlo simulation over a broad range of parameters, including the topology of the interacting network. Thus, epistasis assists the evolutionary transit through evolutionary hurdles leaving marks at the level of haplotype disequilibrium. The method allows determining selection coefficient for each site and the epistasis strength of each pair from a sequence set. The resulting ability to detect clusters of deleterious mutation close to full compensation is essential for biomedical applications. These findings help to understand the role of epistasis in multiple compensatory mutations in viral resistance to antivirals and immune response.

Dominance and Epistasis Interactions Revealed as Important Variants for Leaf Traits of Maize NAM Population

Leaf orientation traits of maize (Zea mays) are complex traits controlling by multiple loci with additive, dominance, epistasis, and environmental interaction effects. In this study, an attempt was made for identifying the causal loci, and estimating the additive, non-additive, environmental specific genetic effects underpinning leaf traits (leaf length, leaf width, and upper leaf angle) of maize NAM population. Leaf traits were analyzed by using full genetic model and additive model of multiple loci. Analysis with full genetic model identified 38∼47 highly significant loci (-log₁₀P_EW > 5), while estimated total heritability were 64.32∼79.06% with large contributions due to dominance and dominance related epistasis effects (16.00∼56.91%). Analysis with additive model obtained smaller total heritability ( hT2 ≙ 18.68∼29.56%) and detected fewer loci (30∼36) as compared to the full genetic model. There were 12 pleiotropic loci identified for the three leaf traits: eight loci for leaf length and leaf width, and four loci for leaf length and leaf angle. Optimal genotype combinations of superior line (SL) and superior hybrid (SH) were predicted for each of the traits under four different environments based on estimated genotypic effects to facilitate maker-assisted selection for the leaf traits.

Dissection of complicate genetic architecture and breeding perspective of cottonseed traits by genome-wide association study

BACKGROUND

Cottonseed is one of the most important raw materials for plant protein, oil and alternative biofuel for diesel engines. Understanding the complex genetic basis of cottonseed traits is requisite for achieving efficient genetic improvement of the traits. However, it is not yet clear about their genetic architecture in genomic level. GWAS has been an effective way to explore genetic basis of quantitative traits in human and many crops. This study aims to dissect genetic mechanism seven cottonseed traits by a GWAS for genetic improvement.

RESULTS

A genome-wide association study (GWAS) based on a full gene model with gene effects as fixed and gene-environment interaction as random, was conducted for protein, oil and 5 fatty acids using 316 accessions and ~ 390 K SNPs. Totally, 124 significant quantitative trait SNPs (QTSs), consisting of 16, 21, 87 for protein, oil and fatty acids (palmitic, linoleic, oleic, myristic, stearic), respectively, were identified and the broad-sense heritability was estimated from 71.62 to 93.43%; no QTS-environment interaction was detected for the protein, the palmitic and the oleic contents; the protein content was predominantly controlled by epistatic effects accounting for 65.18% of the total variation, but the oil content and the fatty acids except the palmitic were mainly determined by gene main effects and no epistasis was detected for the myristic and the stearic. Prediction of superior pure line and hybrid revealed the potential of the QTSs in the improvement of cottonseed traits, and the hybrid could achieve higher or lower genetic values compared with pure lines.

CONCLUSIONS

This study revealed complex genetic architecture of seven cottonseed traits at whole genome-wide by mixed linear model approach; the identified genetic variants and estimated genetic component effects of gene, gene-gene and gene-environment interaction provide cotton geneticist or breeders new knowledge on the genetic mechanism of the traits and the potential molecular breeding design strategy.

Association and cis-mQTL analysis of variants in CHRNA3-A5, CHRNA7, CHRNB2, and CHRNB4 in relation to nicotine dependence in a Chinese Han population

Nicotine dependence (ND) is a worldwide health problem. Numerous genetic studies have demonstrated a significant association of variants in nicotinic acetylcholine receptors (nAChRs) with smoking behaviors. However, most of these studies enrolled only subjects of European or African ancestry. In addition, although an increasing body of evidence implies a causal connection of single-nucleotide polymorphisms (SNPs) and epigenetic regulation of gene expression, few studies of this issue have been reported. In this study, we performed both association and interaction analysis for 67 SNPs in CHRNA3-A5, CHRNA7, CHRNB2, and CHRNB4 with ND in a Chinese Han population (N = 5055). We further analyzed cis-mQTL for the three most significant SNPs and 5580 potential methylation loci within these target gene regions. Our results indicated that the SNPs rs1948 and rs7178270 in CHRNB4 and rs3743075 in CHRNA3 were significantly associated with the Fagerström Test for Nicotine Dependence (FTND) score (p = 6.6 × 10^-5; p = 2.0 × 10^-4, and p = 7.0 × 10^-4, respectively). Haplotype-based association analysis revealed that two major haplotypes, T-G and C-A, formed by rs3743075-rs3743074 in CHRNA3, and other two major haplotypes, A-G-C and G-C-C, formed by rs1948-rs7178270-rs17487223 in CHRNB4, were significantly associated with the FTND score (p ≤ 8.0 × 10^-4). Further, we found evidence for the presence of significant interaction among variants within CHRNA3/B4/A5, CHRNA4/B2/A5, and CHRNA7 in affecting ND, with corresponding p values of 5.8 × 10^-6, 8.0 × 10^-5, and 0.012, respectively. Finally, we identified two CpG sites (CpG_2975 and CpG_3007) in CHRNA3 that are significantly associated with three cis-mQTL SNPs (rs1948, rs7178270, rs3743075) in the CHRNA5/A3/B4 cluster (p ≤ 1.9 × 10^-6), which formed four significant CpG-SNP pairs in our sample. Together, we revealed at least three novel SNPs in CHRNA3 and CHRNB4 to be significantly associated with the FTND score. Further, we showed that these significant variants contribute to ND via two methylated sites, and we demonstrated significant interaction affecting ND among variants in CHRNA5/A3/B4, CHRNA7, and CHRNA4/B2/A5. In sum, these findings provide robust evidence that SNPs in nAChR genes convey a risk of ND in the Chinese Han population.

A Review on Methods for Detecting SNP Interactions in High-Dimensional Genomic Data

In this era of genome-wide association studies (GWAS), the quest for understanding the genetic architecture of complex diseases is rapidly increasing more than ever before. The development of high throughput genotyping and next generation sequencing technologies enables genetic epidemiological analysis of large scale data. These advances have led to the identification of a number of single nucleotide polymorphisms (SNPs) responsible for disease susceptibility. The interactions between SNPs associated with complex diseases are increasingly being explored in the current literature. These interaction studies are mathematically challenging and computationally complex. These challenges have been addressed by a number of data mining and machine learning approaches. This paper reviews the current methods and the related software packages to detect the SNP interactions that contribute to diseases. The issues that need to be considered when developing these models are addressed in this review. The paper also reviews the achievements in data simulation to evaluate the performance of these models. Further, it discusses the future of SNP interaction analysis.

Linking Alzheimer's disease and type 2 diabetes: Novel shared susceptibility genes detected by cFDR approach

BACKGROUND

Both type 2 diabetes (T2D) and Alzheimer's disease (AD) occur commonly in the aging populations and T2D has been considered as an important risk factor for AD. The heritability of both diseases is estimated to be over 50%. However, common pleiotropic single-nucleotide polymorphisms (SNPs)/loci have not been well-defined. The aim of this study is to analyze two large public accessible GWAS datasets to identify novel common genetic loci for T2D and/or AD.

METHODS AND MATERIALS

The recently developed novel conditional false discovery rate (cFDR) approach was used to analyze the summary GWAS datasets from International Genomics of Alzheimer's Project (IGAP) and Diabetes Genetics Replication And Meta-analysis (DIAGRAM) to identify novel susceptibility genes for AD and T2D.

RESULTS

We identified 78 SNPs (including 58 novel SNPs) that were associated with AD in Europeans conditional on T2D (cFDR<0.05). 66 T2D SNPs (including 40 novel SNPs) were identified by conditioning on SNPs association with AD (cFDR<0.05). A conjunction-cFDR (ccFDR) analysis detected 8 pleiotropic SNPs with a significance threshold of ccFDR<0.05 for both AD and T2D, of which 5 SNPs (rs6982393, rs4734295, rs7812465, rs10510109, rs2421016) were novel findings. Furthermore, among the 8 SNPs annotated at 6 different genes, 3 corresponding genes TP53INP1, TOMM40 and C8orf38 were related to mitochondrial dysfunction, critically involved in oxidative stress, which potentially contribute to the etiology of both AD and T2D.

CONCLUSION

Our study provided evidence for shared genetic loci between T2D and AD in European subjects by using cFDR and ccFDR analyses. These results may provide novel insight into the etiology and potential therapeutic targets of T2D and/or AD.

Significant association of the CHRNB3-CHRNA6 gene cluster with nicotine dependence in the Chinese Han population

Although numerous studies have revealed significant associations between variants in the nicotinic acetylcholine receptors (nAChR) subunits and nicotine dependence (ND), only few studies were performed in Chinese subjects. Here, we performed association and interaction analysis for 20 single nucleotide polymorphisms (SNPs) in the CHRNB3-CHRNA6 gene cluster with ND in a Chinese Han population (N = 5,055). We found nominally significant associations for all tested SNPs with ND measured by the Fagerström Test for Nicotine Dependence score; of these, 11 SNPs remained significant after Bonferroni correction for multiple tests (p = 9 × 10⁻⁴~2 × 10⁻³). Further conditional analysis indicated that no other SNP was significantly associated with ND independent of the most-highly significant SNP, rs6474414. Also, our haplotype-based association analysis indicated that each haplotype block was significantly associated with ND (p < 0.01). Further, we provide the first evidence of the genetic interaction of these two genes in affecting ND in this sample with an empirical p-value of 0.0015. Finally, our meta-analysis of samples with Asian and European origins for five SNPs in CHRNB3 showed significant associations with ND, with p-values ranging from 6.86 × 10⁻¹⁴ for rs13280604 to 6.50 × 10⁻⁸ for rs4950^. This represents the first study showing that CHRNB3/A6 are highly associated with ND in a large Chinese Han sample.

Genetic dissection of yield traits in super hybrid rice Xieyou9308 using both unconditional and conditional genome-wide association mapping

With the development and application of super rice breeding, elite rice hybrids with super high-yielding potential have been widely developed in last decades in China. Xieyou9308 is one of the most famous super hybrid rice varieties. To uncover the genetic mechanism of Xieyou9308’s high yield potential, a recombinant inbred line (RIL) population derived from cross of XieqingzaoB and Zhonghui9308 was re-sequenced and investigated on the grain yield (GYD) and its three component traits, number of panicles per plant (NP), number of filled grains per panicle (NFGP), and grain weight (GW). Unconditional and conditional genome-wide association analysis, based on a linear mixed model with epistasis and gene-environment interaction effects, were conducted, using ~0.7 million identified SNPs. There were six, four, seven, and seven QTSs identified for GYD, NP, NFGP, and GW, respectively, with accumulated explanatory heritability varying from 43.06% to 48.36%; additive by environment interactions were detected for GYD, some minor epistases were detected for NP and NFGP. Further, conditional genetic mapping analysis for GYD given its three components revealed several novel QTSs associated with yield than that were suppressed in our unconditional mapping analysis.

A multianalytical approach to evaluate the association of 55 SNPs in 28 genes with obesity risk in North Indian adults

OBJECTIVES

The aim of the study was to investigate the association of 55 SNPs in 28 genes with obesity risk in a North Indian population using a multianalytical approach.

METHODS

Overall, 480 subjects from the North Indian population were studied using strict inclusion/exclusion criteria. SNP Genotyping was carried out by Sequenom Mass ARRAY platform (Sequenom, San Diego, CA) and validated Taqman® allelic discrimination (Applied Biosystems® ). Statistical analyses were performed using SPSS software version 19.0, SNPStats, GMDR software (version 6) and GENEMANIA.

RESULTS

Logistic regression analysis of 55 SNPs revealed significant associations (P < .05) of 49 SNPs with BMI linked obesity risk whereas the remaining 6 SNPs revealed no association (P > .05). The pathway-wise G-score revealed the significant role (P = .0001) of food intake-energy expenditure pathway genes. In CART analysis, the combined genotypes of FTO rs9939609 and TCF7L2 rs7903146 revealed the highest risk for BMI linked obesity. The analysis of the FTO-IRX3 locus revealed high LD and high order gene-gene interactions for BMI linked obesity. The interaction network of all of the associated genes in the present study generated by GENEMANIA revealed direct and indirect connections. In addition, the analysis with centralized obesity revealed that none of the SNPs except for FTO rs17818902 were significantly associated (P < .05).

CONCLUSIONS

In this multi-analytical approach, FTO rs9939609 and IRX3 rs3751723, along with TCF7L2 rs7903146 and TMEM18 rs6548238, emerged as the major SNPs contributing to BMI linked obesity risk in the North Indian population.

Comparing GWAS Results of Complex Traits Using Full Genetic Model and Additive Models for Revealing Genetic Architecture

Most of the genome-wide association studies (GWASs) for human complex diseases have ignored dominance, epistasis and ethnic interactions. We conducted comparative GWASs for total cholesterol using full model and additive models, which illustrate the impacts of the ignoring genetic variants on analysis results and demonstrate how genetic effects of multiple loci could differ across different ethnic groups. There were 15 quantitative trait loci with 13 individual loci and 3 pairs of epistasis loci identified by full model, whereas only 14 loci (9 common loci and 5 different loci) identified by multi-loci additive model. Again, 4 full model detected loci were not detected using multi-loci additive model. PLINK-analysis identified two loci and GCTA-analysis detected only one locus with genome-wide significance. Full model identified three previously reported genes as well as several new genes. Bioinformatics analysis showed some new genes are related with cholesterol related chemicals and/or diseases. Analyses of cholesterol data and simulation studies revealed that the full model performs were better than the additive-model performs in terms of detecting power and unbiased estimations of genetic variants of complex traits.

Association of FTO and IRX3 genetic variants to obesity risk in north India

BACKGROUND

Obesity is an increasingly important health problem worldwide as well as in developing countries like India. Recent genetic studies suggest that obesity associated FTO and IRX3 are functionally linked and many effects due to genetic variants in FTO gene act through IRX3.

AIM

To evaluate the association of FTO and IRX3 genetic variants towards obesity risk.

SUBJECTS AND METHODS

North Indian individuals categorised as non-obese (BMI < 30 kg/m(2)) and obese (BMI ≥ 30 kg/m(2)) were selected. FTO rs8050136, rs1421085, rs9939609, rs17817449 and IRX3 rs3751723 were genotyped by means of validated Taqman® allelic discrimination to evaluate their association with obesity by means of single locus logistic regression by SPSS ver. 19 and multi-locus linkage and haplotype analysis by SNPStats and gene-gene interaction with Generalised Multifactor Dimensionality Reduction (GMDR) ver.6.

RESULTS

In single locus analysis, FTO rs8050136 CA (p = 0.0001; OR (95% CI) = 2.4 (1.7-3.4) and AA (p = 0.0001; OR (95% CI) = 3.1 (1.9-5.2); FTO rs1421085 TA (p = 0.0001; OR (95% CI) = 2.1 (1.4-3.0) and AA (p = 0.0001; OR (95% CI) = 3.0 (1.8-5.0); FTO rs9939609 TC (p = 0.0001; OR (95% CI) = 2.1 (1.5-3.1) and CC (p = 0.0001; OR (95% CI) = 4.2 (2.5-7.3) along with TG (p = 0.001; OR (95% CI) = 2.1 (1.3-3.2) and GG (p = 0.021; OR (95% CI) = 3.8 (1.2-11.8) genotypes of FTO rs17817449 with GT (p = 0.0001; OR (95% CI) = 2.1 (1.5-3.1) and TT (p = 0.012; OR (95% CI) = 3.3 (1.8-3.6) genotypes of IRX3 rs3751723 were significantly associated with obesity. In multi-locus analysis, SNPs of FTO and IRX3 were in strong linkage disequilibrium and in haplotype and GMDR analysis the SNPs were significantly associated with obesity risk (p < 0.05).

CONCLUSION

This is the first study to reveal that genetic variants of both FTO and IRX3 genes are in high linkage disequilibrium (LD) and are associated with obesity risk in North Indians.

Bioinformatics Approach in Plant Genomic Research

The advance in genomics technology leads to the dramatic change in plant biology research. Plant biologists now easily access to enormous genomic data to deeply study plant high-density genetic variation at molecular level. Therefore, fully understanding and well manipulating bioinformatics tools to manage and analyze these data are essential in current plant genome research. Many plant genome databases have been established and continued expanding recently. Meanwhile, analytical methods based on bioinformatics are also well developed in many aspects of plant genomic research including comparative genomic analysis, phylogenomics and evolutionary analysis, and genome-wide association study. However, constantly upgrading in computational infrastructures, such as high capacity data storage and high performing analysis software, is the real challenge for plant genome research. This review paper focuses on challenges and opportunities which knowledge and skills in bioinformatics can bring to plant scientists in present plant genomics era as well as future aspects in critical need for effective tools to facilitate the translation of knowledge from new sequencing data to enhancement of plant productivity.

Dissection of genetic architecture of rice plant height and heading date by multiple-strategy-based association studies

Xieyou9308 is a certified super hybrid rice cultivar with a high grain yield. To investigate its underlying genetic basis of high yield potential, a recombinant inbred line (RIL) population derived from the cross between the maintainer line XieqingzaoB (XQZB) and the restorer line Zhonghui9308 (ZH9308) was constructed for identification of quantitative trait SNPs (QTSs) associated with two important agronomic traits, plant height (PH) and heading date (HD). By re-sequencing of 138 recombinant inbred lines (RILs), a total of ~0.7 million SNPs were identified for the association studies on the PH and HD. Three association mapping strategies (including hypothesis-free genome-wide association and its two complementary hypothesis-engaged ones, QTL-based association and gene-based association) were adopted for data analysis. Using a saturated mixed linear model including epistasis and environmental interaction, we identified a total of 31 QTSs associated with either the PH or the HD. The total estimated heritability across three analyses ranged from 37.22% to 45.63% and from 37.53% to 55.96% for the PH and HD, respectively. In this study we examined the feasibility of association studies in an experimental population (RIL) and identified several common loci through multiple strategies which could be preferred candidates for further research.

Functional Regression Models for Epistasis Analysis of Multiple Quantitative Traits

To date, most genetic analyses of phenotypes have focused on analyzing single traits or analyzing each phenotype independently. However, joint epistasis analysis of multiple complementary traits will increase statistical power and improve our understanding of the complicated genetic structure of the complex diseases. Despite their importance in uncovering the genetic structure of complex traits, the statistical methods for identifying epistasis in multiple phenotypes remains fundamentally unexplored. To fill this gap, we formulate a test for interaction between two genes in multiple quantitative trait analysis as a multiple functional regression (MFRG) in which the genotype functions (genetic variant profiles) are defined as a function of the genomic position of the genetic variants. We use large-scale simulations to calculate Type I error rates for testing interaction between two genes with multiple phenotypes and to compare the power with multivariate pairwise interaction analysis and single trait interaction analysis by a single variate functional regression model. To further evaluate performance, the MFRG for epistasis analysis is applied to five phenotypes of exome sequence data from the NHLBI’s Exome Sequencing Project (ESP) to detect pleiotropic epistasis. A total of 267 pairs of genes that formed a genetic interaction network showed significant evidence of epistasis influencing five traits. The results demonstrate that the joint interaction analysis of multiple phenotypes has a much higher power to detect interaction than the interaction analysis of a single trait and may open a new direction to fully uncovering the genetic structure of multiple phenotypes.

The widely used statistical methods test interaction for single phenotype. However, we often observe pleotropic genetic interaction effects. The simultaneous gene-gene (GxG) interaction analysis of multiple complementary traits will increase statistical power to detect GxG interactions. Although GxG interactions play an important role in uncovering the genetic structure of complex traits, the statistical methods for detecting GxG interactions in multiple phenotypes remains less developed owing to its potential complexity. Therefore, we extend functional regression model from single variate to multivariate for simultaneous GxG interaction analysis of multiple correlated phenotypes. Large-scale simulations are conducted to evaluate Type I error rates for testing interaction between two genes with multiple phenotypes and to compare power with traditional multivariate pair-wise interaction analysis and single trait interaction analysis by a single variate functional regression model. To further evaluate performance, the MFRG for interaction analysis is applied to five phenotypes of exome sequence data from the NHLBI’s Exome Sequencing Project (ESP) to detect pleiotropic GxG interactions. 267 pairs of genes that formed a genetic interaction network showed significant evidence of interactions influencing five traits.

Detection of Epistatic and Gene-Environment Interactions Underlying Three Quality Traits in Rice Using High-Throughput Genome-Wide Data

With development of sequencing technology, dense single nucleotide polymorphisms (SNPs) have been available, enabling uncovering genetic architecture of complex traits by genome-wide association study (GWAS). However, the current GWAS strategy usually ignores epistatic and gene-environment interactions due to absence of appropriate methodology and heavy computational burden. This study proposed a new GWAS strategy by combining the graphics processing unit- (GPU-) based generalized multifactor dimensionality reduction (GMDR) algorithm with mixed linear model approach. The reliability and efficiency of the analytical methods were verified through Monte Carlo simulations, suggesting that a population size of nearly 150 recombinant inbred lines (RILs) had a reasonable resolution for the scenarios considered. Further, a GWAS was conducted with the above two-step strategy to investigate the additive, epistatic, and gene-environment associations between 701,867 SNPs and three important quality traits, gelatinization temperature, amylose content, and gel consistency, in a RIL population with 138 individuals derived from super-hybrid rice Xieyou9308 in two environments. Four significant SNPs were identified with additive, epistatic, and gene-environment interaction effects. Our study showed that the mixed linear model approach combining with the GPU-based GMDR algorithm is a feasible strategy for implementing GWAS to uncover genetic architecture of crop complex traits.

Mixed Linear Model Approaches of Association Mapping for Complex Traits Based on Omics Variants

Precise prediction for genetic architecture of complex traits is impeded by the limited understanding on genetic effects of complex traits, especially on gene-by-gene (GxG) and gene-by-environment (GxE) interaction. In the past decades, an explosion of high throughput technologies enables omics studies at multiple levels (such as genomics, transcriptomics, proteomics, and metabolomics). The analyses of large omics data, especially two-loci interaction analysis, are very time intensive. Integrating the diverse omics data and environmental effects in the analyses also remain challenges. We proposed mixed linear model approaches using GPU (Graphic Processing Unit) computation to simultaneously dissect various genetic effects. Analyses can be performed for estimating genetic main effects, GxG epistasis effects, and GxE environment interaction effects on large-scale omics data for complex traits, and for estimating heritability of specific genetic effects. Both mouse data analyses and Monte Carlo simulations demonstrated that genetic effects and environment interaction effects could be unbiasedly estimated with high statistical power by using the proposed approaches.

An Epistatic Interaction between Themis1 and Vav1 Modulates Regulatory T Cell Function and Inflammatory Bowel Disease Development

The development of inflammatory diseases depends on complex interactions between several genes and various environmental factors. Discovering new genetic risk factors and understanding the mechanisms whereby they influence disease development is of paramount importance. We previously reported that deficiency in Themis1, a new actor of TCR signaling, impairs regulatory T cell (Treg) function and predisposes Brown-Norway (BN) rats to spontaneous inflammatory bowel disease (IBD). In this study, we reveal that the epistasis between Themis1 and Vav1 controls the occurrence of these phenotypes. Indeed, by contrast with BN rats, Themis1 deficiency in Lewis rats neither impairs Treg suppressive functions nor induces pathological manifestations. By using congenic lines on the BN genomic background, we show that the impact of Themis1 deficiency on Treg suppressive functions depends on a 117-kb interval coding for a R63W polymorphism that impacts Vav1 expression and functions. Indeed, the introduction of a 117-kb interval containing the Lewis Vav1-R63 variant restores Treg function and protects Themis1-deficient BN rats from spontaneous IBD development. We further show that Themis1 binds more efficiently to the BN Vav1-W63 variant and is required to stabilize its recruitment to the transmembrane adaptor LAT and to fully promote the activation of Erk kinases. Together, these results highlight the importance of the signaling pathway involving epistasis between Themis1 and Vav1 in the control of Treg suppressive function and susceptibility to IBD development.

A roadmap to multifactor dimensionality reduction methods

Complex diseases are defined to be determined by multiple genetic and environmental factors alone as well as in interactions. To analyze interactions in genetic data, many statistical methods have been suggested, with most of them relying on statistical regression models. Given the known limitations of classical methods, approaches from the machine-learning community have also become attractive. From this latter family, a fast-growing collection of methods emerged that are based on the Multifactor Dimensionality Reduction (MDR) approach. Since its first introduction, MDR has enjoyed great popularity in applications and has been extended and modified multiple times. Based on a literature search, we here provide a systematic and comprehensive overview of these suggested methods. The methods are described in detail, and the availability of implementations is listed. Most recent approaches offer to deal with large-scale data sets and rare variants, which is why we expect these methods to even gain in popularity.

Evaluating methods for modeling epistasis networks with application to head and neck cancer

Epistasis helps to explain how multiple single-nucleotide polymorphisms (SNPs) interact to cause disease. A variety of tools have been developed to detect epistasis. In this article, we explore the strengths and weaknesses of an information theory approach for detecting epistasis and compare it to the logistic regression approach through simulations. We consider several scenarios to simulate the involvement of SNPs in an epistasis network with respect to linkage disequilibrium patterns among them and the presence or absence of main and interaction effects. We conclude that the information theory approach more efficiently detects interaction effects when main effects are absent, whereas, in general, the logistic regression approach is appropriate in all scenarios but results in higher false positives. We compute epistasis networks for SNPs in the FSD1L gene using a two-phase head and neck cancer genome-wide association study involving 2,185 cases and 4,507 controls to demonstrate the practical application of the methods.

Multivariate dimensionality reduction approaches to identify gene-gene and gene-environment interactions underlying multiple complex traits

The elusive but ubiquitous multifactor interactions represent a stumbling block that urgently needs to be removed in searching for determinants involved in human complex diseases. The dimensionality reduction approaches are a promising tool for this task. Many complex diseases exhibit composite syndromes required to be measured in a cluster of clinical traits with varying correlations and/or are inherently longitudinal in nature (changing over time and measured dynamically at multiple time points). A multivariate approach for detecting interactions is thus greatly needed on the purposes of handling a multifaceted phenotype and longitudinal data, as well as improving statistical power for multiple significance testing via a two-stage testing procedure that involves a multivariate analysis for grouped phenotypes followed by univariate analysis for the phenotypes in the significant group(s). In this article, we propose a multivariate extension of generalized multifactor dimensionality reduction (GMDR) based on multivariate generalized linear, multivariate quasi-likelihood and generalized estimating equations models. Simulations and real data analysis for the cohort from the Study of Addiction: Genetics and Environment are performed to investigate the properties and performance of the proposed method, as compared with the univariate method. The results suggest that the proposed multivariate GMDR substantially boosts statistical power.

Detecting epistasis in human complex traits

Genome-wide association studies (GWASs) have become the focus of the statistical analysis of complex traits in humans, successfully shedding light on several aspects of genetic architecture and biological aetiology. Single-nucleotide polymorphisms (SNPs) are usually modelled as having additive, cumulative and independent effects on the phenotype. Although evidently a useful approach, it is often argued that this is not a realistic biological model and that epistasis (that is, the statistical interaction between SNPs) should be included. The purpose of this Review is to summarize recent directions in methodology for detecting epistasis and to discuss evidence of the role of epistasis in human complex trait variation. We also discuss the relevance of epistasis in the context of GWASs and potential hazards in the interpretation of statistical interaction terms.

Analysis of genetic association and epistasis interactions between circadian clock genes and symptom dimensions of bipolar affective disorder

Bipolar affective disorder (BD) is a severe psychiatric disorder characterized by periodic changes in mood from depression to mania. Disruptions of biological rhythms increase risk of mood disorders. Because clinical representation of disease is heterogeneous, homogenous sets of patients are suggested to use in the association analyses. In our study, we aimed to apply previously computed structure of bipolar disorder symptom dimension for analyses of genetic association. We based quantitative trait on: main depression, sleep disturbances, appetite disturbances, excitement and psychotic dimensions consisted of OPCRIT checklist items. We genotyped 42 polymorphisms from circadian clock genes: PER3, ARNTL, CLOCK and TIMELSSS from 511 patients BD (n = 292 women and n = 219 men). As quantitative trait we used clinical dimensions, described above. Genetic associations between alleles and quantitative trait were performed using applied regression models applied in PLINK. In addition, we used the Kruskal-Wallis test to look for associations between genotypes and quantitative trait. During second stage of our analyses, we used multidimensional scaling (multifactor dimensionality reduction) for quantitative trait to compute pairwise epistatic interactions between circadian gene variants. We found association between ARNTL variant rs11022778 main depression (p = 0.00047) and appetite disturbances (p = 0.004). In epistatic interaction analyses, we observed two locus interactions between sleep disturbances (p = 0.007; rs11824092 of ARNTL and rs11932595 of CLOCK) as well as interactions of subdimension in main depression and ARNTL variants (p = 0.0011; rs3789327, rs10766075) and appetite disturbances in depression and ARNTL polymorphism (p = 7 × 10(-4); rs11022778, rs156243).

Association and interaction analyses of 5-HT3 receptor and serotonin transporter genes with alcohol, cocaine, and nicotine dependence using the SAGE data

Previous studies have implicated genes encoding the 5-HT3AB receptors (HTR3A and HTR3B) and the serotonin transporter (SLC6A4), both independently and interactively, in alcohol (AD), cocaine (CD), and nicotine dependence (ND). However, whether these genetic effects also exist in subjects with comorbidities remains largely unknown. We used 1,136 African-American (AA) and 2,428 European-American (EA) subjects from the Study of Addiction: Genetics and Environment (SAGE) to determine associations between 88 genotyped or imputed variants within HTR3A, HTR3B, and SLC6A4 and three types of addictions, which were measured by DSM-IV diagnoses of AD, CD, and ND and the Fagerström Test for Nicotine Dependence (FTND), an independent measure of ND commonly used in tobacco research. Individual SNP-based association analysis revealed a significant association of rs2066713 in SLC6A4 with FTND in AA (β = -1.39; P = 1.6E - 04). Haplotype-based association analysis found one major haplotype formed by SNPs rs3891484 and rs3758987 in HTR3B that was significantly associated with AD in the AA sample, and another major haplotype T-T-G, formed by SNPs rs7118530, rs12221649, and rs2085421 in HTR3A, which showed significant association with FTND in the EA sample. Considering the biologic roles of the three genes and their functional relations, we used the GPU-based Generalized Multifactor Dimensionality Reduction (GMDR-GPU) program to test SNP-by-SNP interactions within the three genes and discovered two- to five-variant models that have significant impacts on AD, CD, ND, or FTND. Interestingly, most of the SNPs included in the genetic interaction model(s) for each addictive phenotype are either overlapped or in high linkage disequilibrium for both AA and EA samples, suggesting these detected variants in HTR3A, HTR3B, and SLC6A4 are interactively contributing to etiology of the three addictive phenotypes examined in this study.

Heterogeneous computing architecture for fast detection of SNP-SNP interactions

Background

The extent of data in a typical genome-wide association study (GWAS) poses considerable computational challenges to software tools for gene-gene interaction discovery. Exhaustive evaluation of all interactions among hundreds of thousands to millions of single nucleotide polymorphisms (SNPs) may require weeks or even months of computation. Massively parallel hardware within a modern Graphic Processing Unit (GPU) and Many Integrated Core (MIC) coprocessors can shorten the run time considerably. While the utility of GPU-based implementations in bioinformatics has been well studied, MIC architecture has been introduced only recently and may provide a number of comparative advantages that have yet to be explored and tested.

Results

We have developed a heterogeneous, GPU and Intel MIC-accelerated software module for SNP-SNP interaction discovery to replace the previously single-threaded computational core in the interactive web-based data exploration program SNPsyn. We report on differences between these two modern massively parallel architectures and their software environments. Their utility resulted in an order of magnitude shorter execution times when compared to the single-threaded CPU implementation. GPU implementation on a single Nvidia Tesla K20 runs twice as fast as that for the MIC architecture-based Xeon Phi P5110 coprocessor, but also requires considerably more programming effort.

Conclusions

General purpose GPUs are a mature platform with large amounts of computing power capable of tackling inherently parallel problems, but can prove demanding for the programmer. On the other hand the new MIC architecture, albeit lacking in performance reduces the programming effort and makes it up with a more general architecture suitable for a wider range of problems.

Analysis pipeline for the epistasis search - statistical versus biological filtering

Gene-gene interactions may contribute to the genetic variation underlying complex traits but have not always been taken fully into account. Statistical analyses that consider gene-gene interaction may increase the power of detecting associations, especially for low-marginal-effect markers, and may explain in part the "missing heritability." Detecting pair-wise and higher-order interactions genome-wide requires enormous computational power. Filtering pipelines increase the computational speed by limiting the number of tests performed. We summarize existing filtering approaches to detect epistasis, after distinguishing the purposes that lead us to search for epistasis. Statistical filtering includes quality control on the basis of single marker statistics to avoid the analysis of bad and least informative data, and limits the search space for finding interactions. Biological filtering includes targeting specific pathways, integrating various databases based on known biological and metabolic pathways, gene function ontology and protein-protein interactions. It is increasingly possible to target single-nucleotide polymorphisms that have defined functions on gene expression, though not belonging to protein-coding genes. Filtering can improve the power of an interaction association study, but also increases the chance of missing important findings.