Detecting, characterizing, and interpreting nonlinear gene-gene interactions using multifactor dimensionality reduction

Interaction between maintenance variables of medical ultrasound scanners through multifactor dimensionality reduction

BACKGROUND

Proper maintenance of electro-medical devices is crucial for the quality of care to patients and the economic performance of healthcare organizations. This research aims to identify the interaction between Ultrasound scanners (US) maintenance variables as a function of maintenance indicators: US in service or decommissioned, excessive number of failures, and failure rate. Knowing those interactions, specific maintenance measures will be developed to improve the reliability of the US.

RESEARCH DESIGN AND METHODS

Multifactor Dimensionality Reduction (MDR) method was eployed to analyze data from 222 US and their four-year maintenance history. Models were developed based on the variables with the greatest influence on maintenance indicators, where US were classified according to the associated risk.

RESULTS

US with more than one major failure or at least one major component replacement had up to 496.4% more failures than the average. Failure rate increased by up to 188.7% over the average for those US with more than three moderate failures, three replacements, or both.

CONCLUSIONS

This study identifies and quantifies the causes of risk to establish a specific maintenance plan for US. It helps to better understand the degradation of US to optimize their operation and maintenance.

Protein interaction networks define the genetic architecture of preterm birth

The likely genetic architecture of complex diseases is that subgroups of patients share variants in genes in specific networks sufficient to express a shared phenotype. We combined high throughput sequencing with advanced bioinformatic approaches to identify such subgroups of patients with variants in shared networks. We performed targeted sequencing of patients with 2 or 3 generations of preterm birth on genes, gene sets and haplotype blocks that were highly associated with preterm birth. We analyzed the data using a multi-sample, protein–protein interaction (PPI) tool to identify significant clusters of patients associated with preterm birth. We identified shared protein interaction networks among preterm cases in two statistically significant clusters, p < 0.001. We also found two small control-dominated clusters. We replicated these data on an independent, large birth cohort. Separation testing showed significant similarity scores between the clusters from the two independent cohorts of patients. Canonical pathway analysis of the unique genes defining these clusters demonstrated enrichment in inflammatory signaling pathways, the glucocorticoid receptor, the insulin receptor, EGF and B-cell signaling, These results support a genetic architecture defined by subgroups of patients that share variants in genes in specific networks and pathways which are sufficient to give rise to the disease phenotype.

Contribution of revision amputation vs replantation for certain digits to functional outcomes after traumatic digit amputations: A comparative study based on multicenter prospective cohort

BACKGROUND

Traumatic digit amputations can result in significant impairment. Optimal surgical treatment is unclear for certain digits in various amputation patterns. Our aim was to compare the contribution of revision amputation vs replantation for each particular digit to functional outcomes.

MATERIALS AND METHODS

Prospective cohort study at three tertiary hospitals was conducted in China. Eligible participants were 3192 patients with traumatic digit amputations enrolled from January 1, 2014, to January 1, 2018. The primary outcome was Michigan Hand Outcomes Questionnaire (MHQ) scores 2 years after initial surgery. Secondary outcome was score on the Disabilities of the Arm, Shoulder, and Hand (DASH).

RESULTS

Of 3192 enrolled patients, 2890 completed the study. Main-effect linear regression showed that participants with replantation of thumb, index, long, and ring (proximal to the proximal interphalangeal [PIP] joint) fingers had significantly better MHQ scores compared to participants with the corresponding finger revision amputation. DASH results were comparable. Finger-finger interaction analyses conducted with multifactor dimensionality reduction (MDR) revealed that the small finger and ring finger had the smallest and greatest interactions with other fingers, respectively. After stratification by amputation level of thumb, index finger, or long finger, linear regression showed that replantation of the ring finger distal to the PIP joint resulted in better MHQ and DASH when the thumb or long finger was also traumatically amputated proximal to the IP/PIP joint.

CONCLUSIONS

Replantation of the thumb, index, long, and ring (proximal to PIP joint) fingers is preferable to revision amputation, regardless of amputation pattern. Replantation of the ring finger amputated distal to PIP was beneficial only when the thumb or long finger was amputated proximal to IP/PIP joint. Replantation or revision amputation of the small finger was indistinguishable in terms of functional outcome. Future investigations and clinical decisions should take into account the role of finger-finger interactions.

Approximate distance correlation for selecting highly interrelated genes across datasets

With the rapid accumulation of biological omics datasets, decoding the underlying relationships of cross-dataset genes becomes an important issue. Previous studies have attempted to identify differentially expressed genes across datasets. However, it is hard for them to detect interrelated ones. Moreover, existing correlation-based algorithms can only measure the relationship between genes within a single dataset or two multi-modal datasets from the same samples. It is still unclear how to quantify the strength of association of the same gene across two biological datasets with different samples. To this end, we propose Approximate Distance Correlation (ADC) to select interrelated genes with statistical significance across two different biological datasets. ADC first obtains the k most correlated genes for each target gene as its approximate observations, and then calculates the distance correlation (DC) for the target gene across two datasets. ADC repeats this process for all genes and then performs the Benjamini-Hochberg adjustment to control the false discovery rate. We demonstrate the effectiveness of ADC with simulation data and four real applications to select highly interrelated genes across two datasets. These four applications including 21 cancer RNA-seq datasets of different tissues; six single-cell RNA-seq (scRNA-seq) datasets of mouse hematopoietic cells across six different cell types along the hematopoietic cell lineage; five scRNA-seq datasets of pancreatic islet cells across five different technologies; coupled single-cell ATAC-seq (scATAC-seq) and scRNA-seq data of peripheral blood mononuclear cells (PBMC). Extensive results demonstrate that ADC is a powerful tool to uncover interrelated genes with strong biological implications and is scalable to large-scale datasets. Moreover, the number of such genes can serve as a metric to measure the similarity between two datasets, which could characterize the relative difference of diverse cell types and technologies.

Association between endothelial nitric oxide synthase and the renin-angiotensin-aldosterone system polymorphisms, blood pressure and training status in normotensive/pre-hypertension and hypertensive older adults: a pilot study

Introduction:Variations in blood pressure (BP) are, in part, genetically determined and some polymorphisms of renin-angiotensin- aldosterone system (RAAS) and synthase of endothelial nitric oxide (eNOS) have been related to hypertension (HT). Conversely, physical exercise is considered a non-pharmacological tool for HT control, treatment, and prevention.Objective: The purpose of this study is to investigate the relationship between eNOS and RAAS polymorphisms, their epistatic interaction, and the respective humoral factors in the BP control in normotensive/pre-hypertension and hypertensive older adults and how this relationship can be modulated by training status (TS) level.Methods:A total of 155 older adults (66.94 ± 6.83 years old) performed the following evaluations: AAHPERD battery test to determine the general functional fitness index (GFFI), systolic and diastolic blood pressure (SBP and DBP), blood collection for DNA extraction, analysis of eNOS gene polymorphisms rs2070744; rs61722009 and rs1799983 and RAAS polymorphisms rs699; rs1799752 and rs5186, and quantification of ACE activity (Fluorimetric Assay) and nitrite concentration (Chemiluminescence Method).Results and Conclusion:Good TS level appears to exert greater influence on SBP for G2 and G3 (G1: 125.79 ± 14.03/ G2: 119.91 ± 11.72^/G3: 119.71 ± 10.85) and on NO₂ for G3 (G1: 0.42 ± 0.25/ G2: 0.54 ± 0.45/ G3: 0.71 ± 0.52). No associations were observed between eNOS and RAAS polymorphisms, but the epistasis was identified between eNOS polymorphism, rs2070744, and RAAS polymorphism, rs699, revealing a statistically significant interaction (p = .0235) with training score of 0.63, a training test accuracy of 0.61 and a cross-validation consistency of 10/10. This result suggests an increased risk of hypertension.

Genetic variants in the cholesterol biosynthesis pathway genes and risk of prostate cancer

Previous studies have found the relationship between cholesterol biosynthesis pathway genes and the risk or prognosis of prostate cancer (PCa), while there is no definite evidence that genetic variants in the cholesterol biosynthesis pathway gene is related to PCa risk. Consequently, we performed this study to explore the associations of single-nucleotide polymorphisms (SNPs) in the cholesterol biosynthesis pathway with PCa risk. We systematically evaluated the association of SNPs in 21 cholesterol biosynthesis pathway genes with the risk of PCa using the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial database using a logistic regression model. Gene expression data of PCa from Gene Expression Omnibus (GEO) datasets and the Cancer Genome Atlas (TCGA) database were applied for mRNA expression analysis. The TCGA database was used to perform expression quantitative trait loci (eQTL) analysis. The interaction between demographic factors and SNPs was analyzed using two-by-four tables. We found T allele of rs67415672 in HMGCS1 is a significant protective allele of PCa [adjusted odds ratio (OR) = 0.90, 95% confidence interval (CI) = 0.83-0.97, P = 4.16 × 10^-3]. Moreover, rs67415672 was an eQTL for HMGCS1 (P = 2.23 × 10^-6). The expression of HMGCS1 significantly decreased in PCa primary tumors than that in normal tissues. These findings indicated that the HMGCS1 rs67415672 might be possible functional susceptibility loci for PCa.

Proteinarium: Multi-sample protein-protein interaction analysis and visualization tool

We posit the likely architecture of complex diseases is that subgroups of patients share variants in genes in specific networks which are sufficient to give rise to a shared phenotype. We developed Proteinarium, a multi-sample protein-protein interaction (PPI) tool, to identify clusters of patients with shared gene networks. Proteinarium converts user defined seed genes to protein symbols and maps them onto the STRING interactome. A PPI network is built for each sample using Dijkstra's algorithm. Pairwise similarity scores are calculated to compare the networks and cluster the samples. A layered graph of PPI networks for the samples in any cluster can be visualized. To test this newly developed analysis pipeline, we reanalyzed publicly available data sets, from which modest outcomes had previously been achieved. We found significant clusters of patients with unique genes which enhanced the findings in the original study.

Epistasis analysis of metabolic genes polymorphisms associated with ischemic heart disease in Yucatan

OBJECTIVE

Epistasis is a type of genetic interaction that could explain much of the phenotypic variability of complex diseases. In this work, the effect of epistasis of metabolic genes and cardiovascular risk on the susceptibility to the development of ischemic heart disease in Yucatan was determined.

METHODS

Case-control study in 79 Yucatecan patients with ischemic heart disease and 101 healthy controls matched by age and origin with cases. The polymorphisms -108CT, Q192R, L55M (paraoxonase 1; PON1), C677T, A1298C (methylenetetrahydrofolate reductase; MTHFR), and the presence/absence of the glutathione S-transferase T1 (GSTT1) gene were genotyped. Epistasis analysis was performed using the multifactorial dimensional reduction method. The best risk prediction model was selected based on precision (%), statistical significance (P<0.05), and cross-validation consistency.

RESULTS

We found an independent association of the null genotype GSTT1*0/0 (OR=3.39, CI: 1.29-8.87, P=0.017) and the null allele (OR=1.86, CI: 1.19-2.91, P=0.007) with ischemic heart disease. The GSTT1*0 deletion and the 677TT genotype (MTHFR) were identified as being at a high cardiovascular risk, whereas the GSTT1*1 wild type genotype and the CC677 variant were at low risk. The gene-environment interaction identified the GSTT1 gene, C677T polymorphism (MTHFR), and hypertension as the factors that best explain ischemic heart disease in the study population.

CONCLUSIONS

The interaction of the MTHFR, GSTT1 and hypertension may constitute a predictive model of risk for early onset ischemic heart disease in the population of Yucatan.

Confident gene activity prediction based on single histone modification H2BK5ac in human cell lines

BACKGROUND

The histones in the core of nucleosomes may be subject to covalent post-transcriptional modifications. These modifications are thought to correlate with and possibly affect various genomic functions, including transcription. Each modification may alone or in combination with other modifications influence or be influenced by transcription. We aimed to identify correlations between single modifications or combinations of modifications at specific nucleosome sized gene regions with transcription activity based on global histone modification and transcription data of human CD4+ T cells and three other human cell lines. Transcription activity was defined in a binary fashion as either on or off. The analysis was done using the Classification and Regression Tree (CART) data mining protocol, and the Multifactorial Dimensionality Reduction (MDR) method was performed to confirm the CART results. These powerful methods have not previously been used for analysis of histone modification data.

RESULTS

We showed that analysis of the single histone modification H2BK5ac at only four gene regions correctly predicted transcription activity status of over 75% of genes in CD4+ T-cells. The H2BK5ac modification status also had high power for prediction of gene transcription activity in the three other cell lines studied. The informative gene regions with the H2BK5ac modification were all positioned proximal to transcription initiation sites. The CART and MDR methods were appropriate tools for the analysis performed. In the study, we also developed a non-arbitrary protocol for binary classification of genes as transcriptionally active or inactive.

CONCLUSIONS

The importance of H2BK5ac modification with regards to transcription control has not previously been emphasized. Analysis of this single modification at only four nucleosome sized gene regions, all of which are at or proximal to transcription initiation, has high power for prediction of gene transcription activity.

A global test for gene-gene interactions based on random matrix theory

Statistical interactions between markers of genetic variation, or gene-gene interactions, are believed to play an important role in the etiology of many multifactorial diseases and other complex phenotypes. Unfortunately, detecting gene-gene interactions is extremely challenging due to the large number of potential interactions and ambiguity regarding marker coding and interaction scale. For many data sets, there is insufficient statistical power to evaluate all candidate gene-gene interactions. In these cases, a global test for gene-gene interactions may be the best option. Global tests have much greater power relative to multiple individual interaction tests and can be used on subsets of the markers as an initial filter prior to testing for specific interactions. In this paper, we describe a novel global test for gene-gene interactions, the global epistasis test (GET), that is based on results from random matrix theory. As we show via simulation studies based on previously proposed models for common diseases including rheumatoid arthritis, type 2 diabetes, and breast cancer, our proposed GET method has superior performance characteristics relative to existing global gene-gene interaction tests. A glaucoma GWAS data set is used to demonstrate the practical utility of the GET method.

Detecting gene-gene interactions using a permutation-based random forest method

Background

Identifying gene-gene interactions is essential to understand disease susceptibility and to detect genetic architectures underlying complex diseases. Here, we aimed at developing a permutation-based methodology relying on a machine learning method, random forest (RF), to detect gene-gene interactions. Our approach called permuted random forest (pRF) which identified the top interacting single nucleotide polymorphism (SNP) pairs by estimating how much the power of a random forest classification model is influenced by removing pairwise interactions.

Results

We systematically tested our approach on a simulation study with datasets possessing various genetic constraints including heritability, number of SNPs, sample size, etc. Our methodology showed high success rates for detecting the interaction SNP pair. We also applied our approach to two bladder cancer datasets, which showed consistent results with well-studied methodologies, such as multifactor dimensionality reduction (MDR) and statistical epistasis network (SEN). Furthermore, we built permuted random forest networks (PRFN), in which we used nodes to represent SNPs and edges to indicate interactions.

Conclusions

We successfully developed a scale-invariant methodology to detect pure gene-gene interactions based on permutation strategies and the machine learning method random forest. This methodology showed great potential to be used for detecting gene-gene interactions to study underlying genetic architectures in a scale-free way, which could be benefit to uncover the complex disease mechanisms.

Epistasis analysis using ReliefF

Here we introduce the ReliefF machine learning algorithm and some of its extensions for detecting and characterizing epistasis in genetic association studies. We provide a general overview of the method and then highlight some of the modifications that have greatly improved its power for genetic analysis. We end with a few examples of published studies of complex human diseases that have used ReliefF.

Complement receptor 1 genetic variants contribute to the susceptibility to gastric cancer in chinese population

As the receptor for C3b/C4b, type 1 complement receptor (CR1/CD35) plays an important role in the regulation of complement activity and is further involved in carcinogenesis. This study aimed to elucidate the association of CR1 genetic variants with the susceptibility to gastric cancer in Chinese population. Based on the NCBI database, totally 13 tag single nucleotide polymorphisms (SNPs) were selected by Haploview program and genotyped using iPlex Gold Genotyping Assay and Sequenom MassArray among 500 gastric cancer cases and 500 healthy controls. Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated by logistic regression to evaluate the association of each SNP with gastric cancer. Of all selected Tag SNPs , CR1 rs9429942 T > C was found to confer to the risk of developing gastric cancer. Compared with the carriers with rs9429942 TT genotype, those with CT genotype had 88% decreased risk of developing gastric cancer with OR (95%CI) of 0.12 (0.03-0.50). Generalized multifactor dimensionality reduction (GMDR) analysis revealed a significant three-way interaction among rs75422544 C > A, rs10494885 C > T and rs7525160 G > C in the development of gastric cancer with a maximum testing balance accuracy of 56.07% and a cross-validation consistency of 7/10 (P = 0.011). In conclusion, our findings demonstrated the genetic role of CR1 gene in the development of gastric cancer in Chinese population.

Heuristic identification of biological architectures for simulating complex hierarchical genetic interactions

Simulation plays an essential role in the development of new computational and statistical methods for the genetic analysis of complex traits. Most simulations start with a statistical model using methods such as linear or logistic regression that specify the relationship between genotype and phenotype. This is appealing due to its simplicity and because these statistical methods are commonly used in genetic analysis. It is our working hypothesis that simulations need to move beyond simple statistical models to more realistically represent the biological complexity of genetic architecture. The goal of the present study was to develop a prototype genotype–phenotype simulation method and software that are capable of simulating complex genetic effects within the context of a hierarchical biology-based framework. Specifically, our goal is to simulate multilocus epistasis or gene–gene interaction where the genetic variants are organized within the framework of one or more genes, their regulatory regions and other regulatory loci. We introduce here the Heuristic Identification of Biological Architectures for simulating Complex Hierarchical Interactions (HIBACHI) method and prototype software for simulating data in this manner. This approach combines a biological hierarchy, a flexible mathematical framework, a liability threshold model for defining disease endpoints, and a heuristic search strategy for identifying high-order epistatic models of disease susceptibility. We provide several simulation examples using genetic models exhibiting independent main effects and three-way epistatic effects.

Interaction between allelic variations in vitamin D receptor and retinoid X receptor genes on metabolic traits

Background

Low vitamin D status has been shown to be a risk factor for several metabolic traits such as obesity, diabetes and cardiovascular disease. The biological actions of 1, 25-dihydroxyvitamin D, are mediated through the vitamin D receptor (VDR), which heterodimerizes with retinoid X receptor, gamma (RXRG). Hence, we examined the potential interactions between the tagging polymorphisms in the VDR (22 tag SNPs) and RXRG (23 tag SNPs) genes on metabolic outcomes such as body mass index, waist circumference, waist-hip ratio (WHR), high- and low-density lipoprotein (LDL) cholesterols, serum triglycerides, systolic and diastolic blood pressures and glycated haemoglobin in the 1958 British Birth Cohort (1958BC, up to n = 5,231). We used Multifactor- dimensionality reduction (MDR) program as a non-parametric test to examine for potential interactions between the VDR and RXRG gene polymorphisms in the 1958BC. We used the data from Northern Finland Birth Cohort 1966 (NFBC66, up to n = 5,316) and Twins UK (up to n = 3,943) to replicate our initial findings from 1958BC.

Results

After Bonferroni correction, the joint-likelihood ratio test suggested interactions on serum triglycerides (4 SNP - SNP pairs), LDL cholesterol (2 SNP - SNP pairs) and WHR (1 SNP - SNP pair) in the 1958BC. MDR permutation model testing analysis showed one two-way and one three-way interaction to be statistically significant on serum triglycerides in the 1958BC. In meta-analysis of results from two replication cohorts (NFBC66 and Twins UK, total n = 8,183), none of the interactions remained after correction for multiple testing (P_interaction >0.17).

Conclusions

Our results did not provide strong evidence for interactions between allelic variations in VDR and RXRG genes on metabolic outcomes; however, further replication studies on large samples are needed to confirm our findings.

Predicting expected progeny difference for marbling score in Angus cattle using artificial neural networks and Bayesian regression models

BACKGROUND

Artificial neural networks (ANN) mimic the function of the human brain and are capable of performing massively parallel computations for data processing and knowledge representation. ANN can capture nonlinear relationships between predictors and responses and can adaptively learn complex functional forms, in particular, for situations where conventional regression models are ineffective. In a previous study, ANN with Bayesian regularization outperformed a benchmark linear model when predicting milk yield in dairy cattle or grain yield of wheat. Although breeding values rely on the assumption of additive inheritance, the predictive capabilities of ANN are of interest from the perspective of their potential to increase the accuracy of prediction of molecular breeding values used for genomic selection. This motivated the present study, in which the aim was to investigate the accuracy of ANN when predicting the expected progeny difference (EPD) of marbling score in Angus cattle. Various ANN architectures were explored, which involved two training algorithms, two types of activation functions, and from 1 to 4 neurons in hidden layers. For comparison, BayesCπ models were used to select a subset of optimal markers (referred to as feature selection), under the assumption of additive inheritance, and then the marker effects were estimated using BayesCπ with π set equal to zero. This procedure is referred to as BayesCpC and was implemented on a high-throughput computing cluster.

RESULTS

The ANN with Bayesian regularization method performed equally well for prediction of EPD as BayesCpC, based on prediction accuracy and sum of squared errors. With the 3K-SNP panel, for example, prediction accuracy was 0.776 using BayesCpC, and ranged from 0.776 to 0.807 using BRANN. With the selected 700-SNP panel, prediction accuracy was 0.863 for BayesCpC and ranged from 0.842 to 0.858 for BRANN. However, prediction accuracy for the ANN with scaled conjugate gradient back-propagation was lower, ranging from 0.653 to 0.689 with the 3K-SNP panel, and from 0.743 to 0.793 with the selected 700-SNP panel.

CONCLUSIONS

ANN with Bayesian regularization performed as well as linear Bayesian regression models in predicting additive genetic values, supporting the idea that ANN are useful as universal approximators of functions of interest in breeding contexts.

Contributory role of five common polymorphisms of RAGE and APE1 genes in lung cancer among Han Chinese

BACKGROUND

Lung cancer is the leading cause of cancer mortality in China. Given the ubiquitous nature of gene-to-gene interaction in lung carcinogenesis, we sought to evaluate five common polymorphisms from advanced glycosylation end product-specific receptor (RAGE) and apurinic/apyrimidinic endonuclease 1 (APE1) genes in association with lung cancer among Han Chinese.

METHODS AND RESULTS

819 patients with lung cancer and 803 cancer-free controls were recruited from Qiqihar city. Genotypes of five examined polymorphisms (RAGE gene: rs1800625, rs1800624, rs2070600; APE1 gene: rs1760944, rs1130409) were determined by ligase detection reaction method. Data were analyzed by R software and multifactor dimensionality reduction (MDR). Hardy-Weinberg equilibrium was satisfied for all five polymorphisms. Overall differences in the genotype and allele distributions were significant for rs1800625 (Pgenotype<0.0005; Pallele<0.0005), rs2070600 (Pgenotype = 0.005; Pallele = 0.004) and rs1130409 (Pgenotype = 0.009; Pallele = 0.004) polymorphisms. Haplotype C-A-A (alleles in order of rs1800625, rs1800624 and rs2070600) of RAGE gene was overrepresented in patients, and conferred a 2.1-fold increased risk of lung cancer (95% confidence interval: 1.52-2.91), independent of confounding factors. Further application of MDR method to five examined polymorphisms identified the overall best interaction model including rs2070600 and rs1130409 polymorphisms. This model had a maximal testing accuracy of 64.63% and a maximal cross-validation consistency of 9 out of 10 at the significant level of 0.006.

CONCLUSIONS

Our findings demonstrated a potential interactive contribution of RAGE and APE1 genes to the pathogenesis of lung cancer among Han Chinese. Further studies are warranted to confirm or refute these findings.

Interactions between genetic variants in AMH and AMHR2 may modify age at natural menopause

The onset of menopause has important implications on women’s fertility and health. We previously identified genetic variants in genes involved in initial follicle recruitment as potential modifiers of age at natural menopause. The objective of this study was to extend our previous study, by searching for pairwise interactions between tagging single nucleotide polymorphisms (tSNPs) in the 5 genes previously selected (AMH, AMHR2, BMP15, FOXL2, GDF9). We performed a cross-sectional study among 3445 women with a natural menopause participating in the Prospect-EPIC study, a population-based prospective cohort study, initiated between 1993 and 1997. Based on the model-based multifactor dimensionality reduction (MB-MDR) test with a permutation-based maxT correction for multiple testing, we found a statistically significant interaction between rs10407022 in AMH and rs11170547 in AMHR2 (p = 0.019) associated with age at natural menopause. Rs10407022 did not have a statistically significant main effect. However, rs10407022 is an eQTL SNP that has been shown to influence mRNA expression levels in lymphoblastoid cell lines. This study provides additional insights into the genetic background of age at natural menopause and suggests a role of the AMH signaling pathway in the onset of natural menopause. However, these results remain suggestive and replication by independent studies is necessary.

Risk score modeling of multiple gene to gene interactions using aggregated-multifactor dimensionality reduction

BACKGROUND

Multifactor Dimensionality Reduction (MDR) has been widely applied to detect gene-gene (GxG) interactions associated with complex diseases. Existing MDR methods summarize disease risk by a dichotomous predisposing model (high-risk/low-risk) from one optimal GxG interaction, which does not take the accumulated effects from multiple GxG interactions into account.

RESULTS

We propose an Aggregated-Multifactor Dimensionality Reduction (A-MDR) method that exhaustively searches for and detects significant GxG interactions to generate an epistasis enriched gene network. An aggregated epistasis enriched risk score, which takes into account multiple GxG interactions simultaneously, replaces the dichotomous predisposing risk variable and provides higher resolution in the quantification of disease susceptibility. We evaluate this new A-MDR approach in a broad range of simulations. Also, we present the results of an application of the A-MDR method to a data set derived from Juvenile Idiopathic Arthritis patients treated with methotrexate (MTX) that revealed several GxG interactions in the folate pathway that were associated with treatment response. The epistasis enriched risk score that pooled information from 82 significant GxG interactions distinguished MTX responders from non-responders with 82% accuracy.

CONCLUSIONS

The proposed A-MDR is innovative in the MDR framework to investigate aggregated effects among GxG interactions. New measures (pOR, pRR and pChi) are proposed to detect multiple GxG interactions.

Pathway-based genetic analysis of preterm birth

Preterm birth in the United States is now 12%. Multiple genes, gene networks, and variants have been associated with this disease. Using a custom database for preterm birth (dbPTB) with a refined set of genes extensively curated from literature and biological databases, we analyzed GWAS of preterm birth for complete genotype data on nearly 2000 preterm and term mothers. We used both the curated genes and a genome-wide approach to carry out a pathway-based analysis. There were 19 significant pathways, which withstood FDR correction for multiple testing that were identified using both the curated genes and the genome-wide approach. The analysis based on the curated genes was more significant than genome-wide in 15 out of 19 pathways. This approach demonstrates the use of a validated set of genes, in the analysis of otherwise unsuccessful GWAS data, to identify gene-gene interactions in a way that enhances statistical power and discovery.

Epistasis, complexity, and multifactor dimensionality reduction

Genome-wide association studies (GWASs) and other high-throughput initiatives have led to an information explosion in human genetics and genetic epidemiology. Conversion of this wealth of new information about genomic variation to knowledge about public health and human biology will depend critically on the complexity of the genotype to phenotype mapping relationship. We review here computational approaches to genetic analysis that embrace, rather than ignore, the complexity of human health. We focus on multifactor dimensionality reduction (MDR) as an approach for modeling one of these complexities: epistasis or gene-gene interaction.

A comparison of internal model validation methods for multifactor dimensionality reduction in the case of genetic heterogeneity

BACKGROUND

Determining the genes responsible for certain human traits can be challenging when the underlying genetic model takes a complicated form such as heterogeneity (in which different genetic models can result in the same trait) or epistasis (in which genes interact with other genes and the environment). Multifactor Dimensionality Reduction (MDR) is a widely used method that effectively detects epistasis; however, it does not perform well in the presence of heterogeneity partly due to its reliance on cross-validation for internal model validation. Cross-validation allows for only one "best" model and is therefore inadequate when more than one model could cause the same trait. We hypothesize that another internal model validation method known as a three-way split will be better at detecting heterogeneity models.

RESULTS

In this study, we test this hypothesis by performing a simulation study to compare the performance of MDR to detect models of heterogeneity with the two different internal model validation techniques. We simulated a range of disease models with both main effects and gene-gene interactions with a range of effect sizes. We assessed the performance of each method using a range of definitions of power.

CONCLUSIONS

Overall, the power of MDR to detect heterogeneity models was relatively poor, especially under more conservative (strict) definitions of power. While the overall power was low, our results show that the cross-validation approach greatly outperformed the three-way split approach in detecting heterogeneity. This would motivate using cross-validation with MDR in studies where heterogeneity might be present. These results also emphasize the challenge of detecting heterogeneity models and the need for further methods development.

Risk estimation and risk prediction using machine-learning methods

After an association between genetic variants and a phenotype has been established, further study goals comprise the classification of patients according to disease risk or the estimation of disease probability. To accomplish this, different statistical methods are required, and specifically machine-learning approaches may offer advantages over classical techniques. In this paper, we describe methods for the construction and evaluation of classification and probability estimation rules. We review the use of machine-learning approaches in this context and explain some of the machine-learning algorithms in detail. Finally, we illustrate the methodology through application to a genome-wide association analysis on rheumatoid arthritis.

Exploring data from genetic association studies using Bayesian variable selection and the Dirichlet process: application to searching for gene × gene patterns

We construct data exploration tools for recognizing important covariate patterns associated with a phenotype, with particular focus on searching for association with gene-gene patterns. To this end, we propose a new variable selection procedure that employs latent selection weights and compare it to an alternative formulation. The selection procedures are implemented in tandem with a Dirichlet process mixture model for the flexible clustering of genetic and epidemiological profiles. We illustrate our approach with the aid of simulated data and the analysis of a real data set from a genome-wide association study.

Genomics and bioinformatics of Parkinson's disease

Within the last two decades, genomics and bioinformatics have profoundly impacted our understanding of the molecular mechanisms of Parkinson's disease (PD). From the description of the first PD gene in 1997 until today, we have witnessed the emergence of new technologies that have revolutionized our concepts to identify genetic mechanisms implicated in human health and disease. Driven by the publication of the human genome sequence and followed by the description of detailed maps for common genetic variability, novel applications to rapidly scrutinize the entire genome in a systematic, cost-effective manner have become a reality. As a consequence, about 30 genetic loci have been unequivocally linked to the pathogenesis of PD highlighting essential molecular pathways underlying this common disorder. Herein we discuss how neurogenomics and bioinformatics are applied to dissect the nature of this complex disease with the overall aim of developing rational therapeutic interventions.

Synergistic association of PTGS2 and CYP2E1 genetic polymorphisms with lung cancer risk in northeastern Chinese

BACKGROUND

Lung cancer is the most common cause of cancer-related deaths worldwide. The aim of this study was to investigate the association of five extensively-studied polymorphisms in PTGS2 (rs689466, rs5275, rs20417) and CYP2E1 (rs2031920, rs6413432) genes with lung cancer risk in a large northeastern Chinese population.

METHODOLOGY/PRINCIPAL FINDINGS

This is a hospital-based case-control study involving 684 patients with lung cancer and 604 cancer-free controls. Genotyping was performed using the PCR-LDR method. Data were analyzed using Haplo.stats and MDR programs. There were significant differences between patients and controls in allele/genotype distributions of rs5275 (P = 0.002/0.003) and rs6413432 (P = 0.037/0.044), as well as in genotype distributions of rs689466 (P = 0.02). The risk for lung cancer associated with the rs5275-C mutant allele was decreased by 60% (95% CI [confidence interval]: 0.21-0.74; P = 0.004) under the recessive model. Carriers of rs689466-G mutant allele had a 28% (95% CI: 0.57-0.92; P = 0.008) reduced risk of developing lung cancer relative to the AA genotype carriers. In haplotype analysis, haplotype G-C-C-T (in order of rs689466, rs5275, rs2031920 and rs6413432) decreased the odds of lung cancer by 28% (95% CI: 0.51-0.93; P = 0.019) after adjusting for confounding factors, whereas haplotype A-T-T-T had 1.49-fold (95% CI: 1.21-1.79; P = 0.012) increased risk for lung cancer. Using MDR method, the overall best model including rs5275, rs689466 and rs6413432 polymorphisms was identified with a maximal testing accuracy of 66.1% and a maximal cross-validation consistency of 10 out of 10 (P = 0.003).

CONCLUSIONS/SIGNIFICANCE

Our findings demonstrated a potentially synergistic association of PTGS2 and CYP2E1 polymorphisms with the underlying cause of lung cancer in northeastern Chinese.

Global tests of P-values for multifactor dimensionality reduction models in selection of optimal number of target genes

Background

Multifactor Dimensionality Reduction (MDR) is a popular and successful data mining method developed to characterize and detect nonlinear complex gene-gene interactions (epistasis) that are associated with disease susceptibility. Because MDR uses a combinatorial search strategy to detect interaction, several filtration techniques have been developed to remove genes (SNPs) that have no interactive effects prior to analysis. However, the cutoff values implemented for these filtration methods are arbitrary, therefore different choices of cutoff values will lead to different selections of genes (SNPs).

Methods

We suggest incorporating a global test of p-values to filtration procedures to identify the optimal number of genes/SNPs for further MDR analysis and demonstrate this approach using a ReliefF filter technique. We compare the performance of different global testing procedures in this context, including the Kolmogorov-Smirnov test, the inverse chi-square test, the inverse normal test, the logit test, the Wilcoxon test and Tippett’s test. Additionally we demonstrate the approach on a real data application with a candidate gene study of drug response in Juvenile Idiopathic Arthritis.

Results

Extensive simulation of correlated p-values show that the inverse chi-square test is the most appropriate approach to be incorporated with the screening approach to determine the optimal number of SNPs for the final MDR analysis. The Kolmogorov-Smirnov test has high inflation of Type I errors when p-values are highly correlated or when p-values peak near the center of histogram. Tippett’s test has very low power when the effect size of GxG interactions is small.

Conclusions

The proposed global tests can serve as a screening approach prior to individual tests to prevent false discovery. Strong power in small sample sizes and well controlled Type I error in absence of GxG interactions make global tests highly recommended in epistasis studies.

Analysis of autosomal genes reveals gene-sex interactions and higher total genetic risk in men with systemic lupus erythematosus

OBJECTIVES

Systemic lupus erythematosus (SLE) is a sexually dimorphic autoimmune disease which is more common in women, but affected men often experience a more severe disease. The genetic basis of sexual dimorphism in SLE is not clearly defined. A study was undertaken to examine sex-specific genetic effects among SLE susceptibility loci.

METHODS

A total of 18 autosomal genetic susceptibility loci for SLE were genotyped in a large set of patients with SLE and controls of European descent, consisting of 5932 female and 1495 male samples. Sex-specific genetic association analyses were performed. The sex-gene interaction was further validated using parametric and non-parametric methods. Aggregate differences in sex-specific genetic risk were examined by calculating a cumulative genetic risk score for SLE in each individual and comparing the average genetic risk between male and female patients.

RESULTS

A significantly higher cumulative genetic risk for SLE was observed in men than in women. (P=4.52x10-8) A significant sex-gene interaction was seen primarily in the human leucocyte antigen (HLA) region but also in IRF5, whereby men with SLE possess a significantly higher frequency of risk alleles than women. The genetic effect observed in KIAA1542 is specific to women with SLE and does not seem to have a role in men.

CONCLUSIONS

The data indicate that men require a higher cumulative genetic load than women to develop SLE. These observations suggest that sex bias in autoimmunity could be influenced by autosomal genetic susceptibility loci.

A bioinformatics approach to preterm birth

A vast body of literature has suggested genetic programming of preterm birth. However, there is a complete lack of an organized analysis and stratification of genetic variants that may indeed be involved in the pathogenesis of preterm birth. We developed a novel bioinformatics approach to identify the nominal genetic variants associated with preterm birth. We used semantic data mining to extract all published articles related to preterm birth. Genes identified from public databases and archives of expression arrays were aggregated with genes curated from the literature. Pathway analysis was used to impute genes from pathways identified in the curations. The curated articles and collected genetic information are available in a web-based tool, the database for preterm birth (dbPTB) that forms a unique resource for investigators interested in preterm birth.

Gene-gene and gene-environmental interactions of childhood asthma: a multifactor dimension reduction approach

Background

The importance of gene-gene and gene-environment interactions on asthma is well documented in literature, but a systematic analysis on the interaction between various genetic and environmental factors is still lacking.

Methodology/Principal Findings

We conducted a population-based, case-control study comprised of seventh-grade children from 14 Taiwanese communities. A total of 235 asthmatic cases and 1,310 non-asthmatic controls were selected for DNA collection and genotyping. We examined the gene-gene and gene-environment interactions between 17 single-nucleotide polymorphisms in antioxidative, inflammatory and obesity-related genes, and childhood asthma. Environmental exposures and disease status were obtained from parental questionnaires. The model-free and non-parametrical multifactor dimensionality reduction (MDR) method was used for the analysis. A three-way gene-gene interaction was elucidated between the gene coding glutathione S-transferase P (GSTP1), the gene coding interleukin-4 receptor alpha chain (IL4Ra) and the gene coding insulin induced gene 2 (INSIG2) on the risk of lifetime asthma. The testing-balanced accuracy on asthma was 57.83% with a cross-validation consistency of 10 out of 10. The interaction of preterm birth and indoor dampness had the highest training-balanced accuracy at 59.09%. Indoor dampness also interacted with many genes, including IL13, beta-2 adrenergic receptor (ADRB2), signal transducer and activator of transcription 6 (STAT6). We also used likelihood ratio tests for interaction and chi-square tests to validate our results and all tests showed statistical significance.

Conclusions/Significance

The results of this study suggest that GSTP1, INSIG2 and IL4Ra may influence the lifetime asthma susceptibility through gene-gene interactions in schoolchildren. Home dampness combined with each one of the genes STAT6, IL13 and ADRB2 could raise the asthma risk.

dbPTB: a database for preterm birth

Genome-wide association studies (GWAS) query the entire genome in a hypothesis-free, unbiased manner. Since they have the potential for identifying novel genetic variants, they have become a very popular approach to the investigation of complex diseases. Nonetheless, since the success of the GWAS approach varies widely, the identification of genetic variants for complex diseases remains a difficult problem. We developed a novel bioinformatics approach to identify the nominal genetic variants associated with complex diseases. To test the feasibility of our approach, we developed a web-based aggregation tool to organize the genes, genetic variations and pathways involved in preterm birth. We used semantic data mining to extract all published articles related to preterm birth. All articles were reviewed by a team of curators. Genes identified from public databases and archives of expression arrays were aggregated with genes curated from the literature. Pathway analysis was used to impute genes from pathways identified in the curations. The curated articles and collected genetic information form a unique resource for investigators interested in preterm birth. The Database for Preterm Birth exemplifies an approach that is generalizable to other disorders for which there is evidence of significant genetic contributions.

Evidence for gene-gene epistatic interactions among susceptibility loci for systemic lupus erythematosus

OBJECTIVE

Several confirmed genetic susceptibility loci for lupus have been described. To date, no clear evidence for genetic epistasis in lupus has been established. The aim of this study was to test for gene-gene interactions in a number of known lupus susceptibility loci.

METHODS

Eighteen single-nucleotide polymorphisms tagging independent and confirmed lupus susceptibility loci were genotyped in a set of 4,248 patients with lupus and 3,818 normal healthy control subjects of European descent. Epistasis was tested by a 2-step approach using both parametric and nonparametric methods. The false discovery rate (FDR) method was used to correct for multiple testing.

RESULTS

We detected and confirmed gene-gene interactions between the HLA region and CTLA4, IRF5, and ITGAM and between PDCD1 and IL21 in patients with lupus. The most significant interaction detected by parametric analysis was between rs3131379 in the HLA region and rs231775 in CTLA4 (interaction odds ratio 1.19, Z = 3.95, P = 7.8 × 10(-5) [FDR ≤0.05], P for multifactor dimensionality reduction = 5.9 × 10(-45)). Importantly, our data suggest that in patients with lupus, the presence of the HLA lupus risk alleles in rs1270942 and rs3131379 increases the odds of also carrying the lupus risk allele in IRF5 (rs2070197) by 17% and 16%, respectively (P = 0.0028 and P = 0.0047, respectively).

CONCLUSION

We provide evidence for gene-gene epistasis in systemic lupus erythematosus. These findings support a role for genetic interaction contributing to the complexity of lupus heritability.

Evidence for epistatic gene interactions between growth factor genes in stroke outcome

BACKGROUND AND PURPOSE

Growth factors are thought to modulate neurological function in stroke recovery through effects in angiogenesis, neurogenesis, and neuroprotection.

METHODS

We tested the association of variants in the brain-derived neurotrophic factor (BDNF), fibroblast growth factor 2 (FGF2) and vascular endothelial growth factor A (VEGFA) genes, and epistatic interactions between them, with functional outcome in a sample of 546 stroke patients.

RESULTS

While none of the tested genes was independently associated with stroke outcome, two significant gene-gene interaction models were identified. One model combined one BDNF and three FGF2 markers, with a global odds ratio (OR) (95% confidence interval [CI]) of 4.15 [2.86-6.04]. The second model included one FGF2 and two VEGFA markers with a global OR [95% CI] = 2.54 [1.76-3.67].

CONCLUSIONS

The results provide evidence for gene interactions in stroke outcome, highlighting the complexity of the recovery mechanisms after a stroke event.

An R package implementation of multifactor dimensionality reduction

Background

A breadth of high-dimensional data is now available with unprecedented numbers of genetic markers and data-mining approaches to variable selection are increasingly being utilized to uncover associations, including potential gene-gene and gene-environment interactions. One of the most commonly used data-mining methods for case-control data is Multifactor Dimensionality Reduction (MDR), which has displayed success in both simulations and real data applications. Additional software applications in alternative programming languages can improve the availability and usefulness of the method for a broader range of users.

Results

We introduce a package for the R statistical language to implement the Multifactor Dimensionality Reduction (MDR) method for nonparametric variable selection of interactions. This package is designed to provide an alternative implementation for R users, with great flexibility and utility for both data analysis and research. The 'MDR' package is freely available online at http://www.r-project.org/. We also provide data examples to illustrate the use and functionality of the package.

Conclusions

MDR is a frequently-used data-mining method to identify potential gene-gene interactions, and alternative implementations will further increase this usage. We introduce a flexible software package for R users.