• Users Online: 77
  • Print this page
  • Email this page

 Table of Contents  
Year : 2020  |  Volume : 4  |  Issue : 1  |  Page : 8-15

Rheumatoid arthritis identification using epistasis analysis through computational models

1 Department of Computer Applications and Information Technology, Arignar Anna Government Arts College, Villupuram, Tamil Nadu, India
2 Research Scholar, Computer Science, Arignar Anna Government Arts College, Villupuram, Tamil Nadu, India

Date of Submission28-Nov-2019
Date of Acceptance07-Dec-2019
Date of Web Publication17-Mar-2020

Correspondence Address:
Dr. R Manavalan
Assistant Professor in Computer Science, o/o Head, Department of Computer Applications and Information Technology, Arignar Anna Government Arts College, Villupuram - 605 602, Tamil Nadu
Login to access the Email id

Source of Support: None, Conflict of Interest: None

DOI: 10.4103/bbrj.bbrj_147_19

Rights and Permissions

Rheumatoid arthritis (RA) is an autoimmune disorder that damages joints irreversibly. Many RA illnesses were also related to complex genetic characteristics and genetic interactions as well. Genome-wide association studies (GWASs) analyzing the fundamental RA-related genetic factors over the past two decades. Nonlinear interaction recognition, also known as epistasis identification, plays a crucial part in identifying RA's genetic causes. GWAS recognizes all single nucleotide polymorphisms (SNPs) genetic variants and the interactions between SNPs to identify RA susceptibility. Manual evaluation and interactions of many SNPs were too complicated for physicians. The main objective of this study is to explore various techniques of statistical, machine learning, optimization, so far applied to identify epistasis effect related to arthritis. The challenges behind the computational model and the experimental outcome of various methods were also focused.

Keywords: Epistasis, gene, genetic variations, genome-wide association studies, interactions, rheumatoid arthritis, single nucleotide polymorphisms

How to cite this article:
Manavalan R, Priya S. Rheumatoid arthritis identification using epistasis analysis through computational models. Biomed Biotechnol Res J 2020;4:8-15

How to cite this URL:
Manavalan R, Priya S. Rheumatoid arthritis identification using epistasis analysis through computational models. Biomed Biotechnol Res J [serial online] 2020 [cited 2022 Aug 8];4:8-15. Available from: https://www.bmbtrj.org/text.asp?2020/4/1/8/280862

  Introduction Top

Rheumatoid arthritis (RA) is one of the chronic autoimmune diseases, which lead to joints problem and bone erosion. RA typically influences the immune system by distressing the tissues and causes fluctuating degrees of pain in the wrist, hand, and legs. Causes such as reduced joint mobility, impaired muscle strength, minimal aerobic capacity, and fatigue are identified as contributing factors to decreased physical activities in RA cases, thereby affecting the patient's life quality.[1]

In the last decades, technological progress generated enormous biological data such as genome, different genetic markers, and phenotype information. It is essential to mine susceptible genes and loci that affect phenotypic traits.[2] Genome-wide association study (GWAS) tests millions of single nucleotide polymorphisms (SNPs) and phenotypes to examine biomarkers of RA in human populations.[3] The Genetic Analysis Workshop 15 analyzed RA-associated SNP interactions with several datasets.[1] Researchers identified that the gene did not function on its own but interacted with another gene to alter the phenotype. Bateson used the word epistasis in 1909. It is portrayed as one genetic variation masking effect of another gene variant.[4] GWAS examines factors of genetic risk that has a significant impact on many autoimmune diseases such as RA to a large extent. Despite technological advances, genetic interactions or epistasis was still difficult to identify using both laboratory and statistical techniques.[1] The success rate of most biological experiments was low for the identification of phenotypes impacts. It took time to analyze and identify the statistical effect of millions of SNPs. Computational models were suggested to identify epistasis impacts in relation to RA. This research relies primarily on the analysis of different computing models used to detect epistasis associated to RA.

  Epistasis Interaction for Rheumatoid Arthritis Identification Using Computational Models Top

To address the issues to identify the gene gene interactions (GGIs) effects for RA, the electronic databases such as IEEE Xplore, Scopus, ScienceDirect, Springer Link, Wiley, and Google Scholar were extensively searched for potential studies of GGIs computational models from 2007 to 2019. The reliable online articles such as thesis and book chapters were also searched for literature survey. The most prominent factors related to RA are the genetic variants and environmental causes. The epistasis effect is mainly used for the determination of risk variables connected with RA. A study publicized the statistical report that up to 1% of the world's adult population is influenced by RA.[5] Therefore, the identification of RA-related genetic interactions is to be needed for increasing the life time of patient. The survey of various computational, statistical, and optimization models to recognize the susceptible genetic factors related to RA is portrayed.

  Review on Computational Approaches Involved in Epistasis Detection for Rheumatoid Arthritis Top

In 2007, Julià et al.[6] implemented CARRIE reverse engineering and MDR techniques to identify SNP association in microarray gene expression data from synovial fibroblasts (SFs). The MDR revealed that there was substantial association between the SNPs rs1800797 and rs1290754 and 13 identified genes related to SF reaction to RA pro-inflammatory stimulus.

In 2007, Meng et al.[7] proposed two-staged approach to detect RA disease-related SNPs. Random forests (RFs) evaluated the genetic data in the first phase to discover the important variable. Causal minimum message length was used to design Bayesian networks as complex etiological network to identify the interactions. There are 750 unrelated RA instances and 750 controls in the simulated dataset. The average prediction error (PE) for testing dataset was 12.4%.

In 2007, Ritchie et al.[8] applied MDR and grammatical evolution neural network (GENN) to detect genetic interaction in RA. The proposed approaches were tested over three datasets such as dataset 1 obtained from Carlton et al.,[9] dataset 2 acquired from Plenge et al.[10] and dataset 3 contains 460 RA cases and 46 controls with 2300 SNPs. In both dataset 1 and dataset 2, the SNP rs2476601 in the gene PTPN22 was identified as a best disease locus model. MDR with BA identified SNP_85 as a best model in dataset 3, whereas no correlations are observed by MDR with normalized mutual information (NMI) and GENN.

The canonical correlation analysis (CCA) generally used to recognize only linear relationship. To identify nonlinear linkage among the genes, Yuan et al.[11] proposed Kernel CCA for genetic interactions. Three genes such as C5, VEGFA, and ITGAV were identified with co-association from the RA dataset with 868 patients and 1194 normal people. In 2009, Clarke et al.[12] performed comparative analysis of logistic models, multinominal models, and proportional odds models to detect interacting genes in case–control designs. The experiment is conducted over RA dataset with 868 RA cases and 1194 normal individuals. The most significant epistasis effects found between the gene PADI4 and CTLA4. The logistic models found 255 interactions, while multinomial and proportional odds models identified 263 and 191 interactions, respectively.

In 2009, Buil et al.[13] presented a gene-based approach to find genetic interactions in RA cases. SNP-based test identified 213 significant SNPs and 60 statistically significant SNPs were found by gene-based tests. SNP test identified the SNP rs2476601 in PTPN22 gene related to RA and confirmed with the previous findings.[3],[4] However, gene-based model did not found the SNPs but identified a region linked to the RA at 9q33.2.

In 2009, Greene et al.[14] introduced Spatially Uniform ReliefF (SURF) to detect GGIs. The success rate was assessed with the heritability of 0.1 using 1600 samples. The outcome revealed that SURF's output was much better than ReliefF. The outcome showed that performance of SURF was much greater than ReliefF. Similarly, the selection of SNP in epistasis using the combination of SURF and tuned relieff (TuRF) outperformed the individual TuRF.

In 2009, Yang et al.[15] proposed SNPHarvester method to detect epistasis in GWAS. SNPHarvester recognizes more than two interacting SNPs at any time. The results revealed that three RA-related SNP markers such as rs5029939, rs5029938, and rs582757 in the gene TNFAIP3 increase the RA susceptibility.

In 2009, D'Angelo et al.[16] proposed the least absolute shrinkage and selection operator (LASSO) with principal-component analysis (PCA) to identify GGIs in GWAS. The LASSO-PCA was assessed on RA dataset, and its results were contrasted with LASSO–SNP. For the experimentation, 135 SNPs from 28 genes were included for significant interaction selection. Both techniques recognized two important interactions, such as HLA-DRA × HLA-DRB9 and HLA-DRA × MICA, respectively.

In 2009, Li et al.[17] proposed the PC approach to identify SNP–SNP interaction in RA cases and compared it with the two-stage approach. PC marked many interactions in the HLA region where the strongest interactions with q < 0.001 included TNF-PC3, VEGFA-PC1, and HLA-C-PC1 as well as TNF-PC2, whereas no significant interactions were recognized by the two-stage method.

In 2009, Liang et al.[18] compared MDR, RF, and Omnibus approaches to discover RA-related genetic effects. The proposed approaches were applied over GAW16 data of 868 RA patients and 1194 healthy people with 545,080 SNPs. MDR has established the best 2-locus combinations of TRAF1-C5 and PTPN22 with a prediction accuracy of 57.64%. Such two genes are independently found by RF and Omnibus with significant P values.

In 2010, Briggs et al.[19] proposed a multistage model incorporated with RF and logistic regression to detect epistatic risk factor associated with PTPN22 for RA. The proposed multistage model comprised four stages. In data reduction stage, candidate choromosomal regions were identified for 292 sibling pairs. Stage 2 is known as extension analysis that tests the region of PTPN22 1858T in 677 patients and 750 controls through logistic regression. In replication analysis (stage 3), NARAC II dataset with 947 patients and 1756 controls were tested to identify the risk variants associated with PTPN22 1858T. A pool of 1624 instances and 2506 controls was evaluated in the final phase, known as a combined analysis. The multistage model recognized seven replicating interactions. The SNP variants CEP72, MYO3A, CDH13, and WFDC1 interact with PTPN22 and increase the disease susceptibility of RA.

The canonical correlation-based U statistics (CCU) was recommended by Peng et al.[20] in 2010, for identification of gene co-association. CCU was evaluated over heroin addiction data comprised of 91 cases and 245 controls and RA dataset. The effectiveness of CCU was compared to MDR, LD-based statistics, and LR. CCU recognized OPRD1 and OPRM1 genetic interaction from heroin addiction data; MDR and LD statistics confirmed the same findings. For RA dataset, the co-association between the genes C5-PADI4, C5-PTPN22, and VEGFA-PADI4 was recognized by the CCU and LD statistics. The LR technique recognized only one VEGFA-PADI4 interaction.

In 2010, Beretta et al.[21] introduced survival dimensionality reduction (SDR) to identify epistasis effects. The RA datasets contains 386 cases with anti-tumor necrosis factor (TNF). In active RA cases, SDR discovered an important interaction between rs1801274 and rs10954213SDR linked to anti-TNF therapy.

In 2010, Yang et al.[22] proposed an adaptive group LASSO (AGL) model to detect GGI in GWAS. The application of AGL model was tested in comparison to LASSO and BEAM using WTCCC RA datasets. The BEAM is not recognized any interaction in the RA dataset. RA-related SNP rs6679677 identified by WTCCC was not recognized by AGL. The AGL technique, however, recognized two other SNPs rs948620 and rs7551793, which are close to SNP rs6679677.

In 2010, Wang et al.[23] framed AntEpiSeeker to identify GGIs in case–control studies. The score function chosen for AntEpiSeeker based on Chi-square values to describe the correlation between phenotypes and SNPs. RA dataset comprised 332,831 SNPs of 3503 people. Moreover, the same is applied to assess AntEpiSeeker effectiveness compared to BEAM, SNPHarvester, and the generic ACO method. AntEpiSeeker recognized significant RA-related SNPs and took 5 days to evaluate the dataset, while it took 2 weeks for SNPHarvester.

In 2011, Chanda et al.[24] developed an information theoretic model for identifying gene environment interaction known as AMBROSIA using KWII (K th Way Information Interaction). For RA dataset, AMBROSIA identified the diseased risk combinations such as {C6_153, Age, RA}, {Sex, RA}, {Age, RA}, {C11_389, RA}, {Smoking, RA}, and {C6_153, RA}.

In 2011, Yoshida et al.[25] proposed SNPInterForest to detect genetic interactions and compared with BOOST. With respect to the RA data set, SNPInterForest identified two interacting SNP pairs such as rs17665418–rs2121526 and rs17665418–rs4799934 from the gene PROK2, PCDH15, and BRUNOL4. The examination of an entire RA dataset took approximately 98 h. It took 11 min and 5 h, respectively, for the BOOST and SNPInterForest methods.

In 2013, Dai et al.[26] proposed an aggregated-MDR (A-MDR) method to find GGIs. The GGIs measures such as predisposing odds ratio, predisposing relative risk, predisposing Chi-square were also introduced. The proposed approach was evaluated over methotrexate-treated patients with juvenile idiopathic arthritis (JIA). A-MDR identified 15 SNP pairs in JIA dataset. A-MDR found that the genes such as SLC25A32, ATIC, and MTHFD2 highly influence the genetic interaction.

To analyze two-locus interactions in RA data set, Walters et al.[27] in 2014 developed Epi2 Loc package using R tool. The NOIA statistical model recognized interactions between rs1290754 and rs1800797 with a P = 0.001.

Walters et al.[28] created the Epi2Pen Package in 2014 to examine two-locus relationships in RA. The statistical model NOIA identified an association between rs1290754 and rs1800797, with the P = 0.001.

Two summarizing scores such as Z-sum score (ZSS) and principle-component score (PCS) with scaled sum score (SSS) for epistasis identification were created by Sengupta Chattopadhyay et al.[29] in 2014. The performance methods were also assessed with 868 RA patients and 1194 healthy people. PCS and ZSS discovered the SNP interaction rs7745656 and rs9275572.

In 2015, Li et al.[30] suggested nonparametric gene-based information gain method (GBIGM) for the identification of nonlinear genetic association. In RA dataset, five significant genetic interactions were detected by GBIGM.

In 2016, Mathieu Emily proposed AGGrEGATOr to detect SNP-SNP interactions and its performance was compared to six methods such as CCA-, KCCA-, PCA-based method, composite linkage disequilibrium), partial least square path modeling, and GBIGM. AGGrEGATOr's type I error rate was at the nominal level for the simulated dataset. In comparison with other methods, AGGrEGATOr recognized 13 significant pairs with small P values in RA dataset. The most significant epistatic effect found between the genes STAT4 and C5.[31]

In 2017, Abo Alchamlat et al.[32] suggested an MDR strategy with K-nearest neighbors (KNN-MDR) for gene–gene interaction identification. MDR used majority voting among all groups of people, whereas KNN-MDR only used majority voting within a group of KNN. To measure the efficiency, simulated dataset and RA genotype data comprising 1999 patients and 1504 controls were used. The proposed method was compared with MDR, BOOST, MegaSNPHunter, and AntEpiSeeker. Compared to existing approaches, KNN-MDR accomplished high power for simulated data set. Ten SNP pairs with a balanced accuracy of 89% are recognized from the RA dataset.

In 2018, Sinoquet and Niel [33] developed SMMB and also hybrid SMMB with ACO approach (SMMB-ACO) to detect GGIs. The proposed models were evaluated over simulated and RA datasets relative to BEAM, DASSO-MB, and AntEpiSeeker. In the simulated dataset, SMMB-ACO showed better power compared to others. In RA dataset, SMMB-ACO required 19 h to examine the whole genome, while SMMB, BEAM, DASSO-MB, and AntEpiSeeker needed 23, 53, 17, and 47 h, respectively.

In 2019, Guan et al.[34] proposed DESeeker, a two-stage technique for epistasis identification. Differential evolution's (DE) fundamental steps were used in the first phase to search for SNP combinations. Hill Climbing Local Search is implemented in the second phase for exhaustive scan SNPs. DESeeker was applied over simulated and RA datasets. The effectiveness of DESeeker was compared over BOOST, SNPRuler, AntEpiSeeker, DE, and IEACO. Dataset RA consists of 1999 patients and 1504 healthy population with 332,831 SNPs. The SNP combinations with the high disease risks such as rs2748666 and rs4810485 was found by DESeeker and AntEpiSeeker.

From the above study, the evaluation of different techniques used to detect RA-related GGIs along with the experimental outcome with performance measures is shortly tabulated in [Table 1].
Table 1: The review of various methods used to detect gene–gene interactions related to rheumatoid arthritis

Click here to view

Issues related to detect epistasis interactions

The identification of susceptible genes, gene location, and SNP interactions plays a crucial role for RA patients. Identifying epistasis impacts in RA by finding biological significance is highly overhead for physicians, so the designs of novel computational models are needed. From the survey, the issues faced by the computational methods while identifying epistasis impacts and analyzing the same are given hereunder.

  • A main role in GGIs is selection of necessary variables. It is cumbersome to choose informative variables from the huge dataset. Selecting huge quantities of variables to define the impact of epistasis as well as identifying the role between them increases computational burden
  • The multi-locus assessment for identifying the impacts of epistasis was a critical job since most methods focused only on pair wise interactions
  • The computational burden highly increases while examine the different combinations of genotype to define the risk factors for RA
  • Instead of exploring the large genome sequencing of RA-related SNPs, most research work concentrated only on certain RA-related target genes
  • The relationship between different GGIs and the region susceptible to RA is difficult to describe
  • Inferring the biological significance from the computational model was still not effectively analyzed.

  Performance Measures Top

The performance measure is a significant and computable metric to access GGIs models performance. Different metrics such as power, Type I error rate, odds ratio, cross-validation consistency, sensitivity, PE, precision, balanced accuracy, true positive rate, and area under the curve for diagnosing RA illnesses through epistasis had been provided in the literature. The formulas of various metrics adapted for the diagnoses of RA illness through GGIs models are exposed in [Table 2].
Table 2: Gene–gene interactions performance measures

Click here to view

  Summary Top

Human genetics has established a growing number of genetic associations linked to RA in recent years. Genetic interaction and SNP–SNP interaction play a key part in the study of genetic variants that could lead to RA disease susceptibility. The main contribution of this review of literature study is to identify the genetic interactions and the main effects of the genes related to RA disease susceptibility through computational models. The extensive review articles were collected from various sources from 2007 to June 2019. From this review, it was observed that the candidate genes such as IL6, PTPN22, ITGAV, C5, PADI4, HLA-DQA2, HLA-RA, HLA-DRB9, HLA-A, HLA-B, HLA-C, HLA-DRB5, HLA-DQB1, TRAF1, VEGFA, TNF-PC3, TN-PC2, CD40, CTLA4, and TNFAIP3 were found by various computational models and play a major causes in genetic interactions. Many researchers chose the SNP test from the above-mentioned genes. From the survey, it was found that most of the Epistasis computational models identified that the candidate genes such as PTPN22, HLA x DRA, HLA x DRB, PADI4 and TRAF1 highly influence the RA disease susceptibility. The SNP rs2476601 from the gene PTPN22 plays a significant role in SNP interactions and associated with RA pathology. Similarly, the SNP rs3761847 from TRAF1 gene interacts with other genes and influences RA disease. From the review, the SNPs and candidate genes chosen to diagnose RA through genetic interactions as well as the roles of the candidate genes are exposed in [Table 3].
Table 3: Rheumatoid arthritis candidates genes and its functions

Click here to view

  Conclusion Top

Through identification of susceptible genetic markers in epistasis, a physician can diagnose and prognosis the RA disease to extend the life time of patient. It was cumbersome to manually examine the primary impacts of a gene, interaction effects, and detection of epistasis interaction in different loci. This article therefore explored different statistical and machine learning models engaged in the epistasis identification of RA as well as analyzed the same carefully to find the issues faced by the methods. It was evident from this study that the machine learning approaches are superior in GGIs detection than statistical methods. However, most of the models did not perform exhaustive and complete scan of SNPs in genome-wide and their performance also differs according to different illnesses. Therefore, the stronger hybrid models with machine learning, soft computing, and evolutionary techniques need to be found to manage huge numbers of SNPs and to find genetic causes of RA by epistasis. The issues identified and discussed in this survey promote researchers to contribute efficient epistasis computational model for identifying susceptible RA-related genes.

Financial support and sponsorship


Conflicts of interest

There are no conflicts of interest.

  References Top

Hanaoka BY, Ithurburn MP, Rigsbee CA, Bridges SL Jr, Moellering DR, Gower B, et al. Chronic inflammation in rheumatoid arthritis and mediators of skeletal muscle pathology and physical impairment: A review. Arthritis Care Res (Hoboken) 2019;71:173-7.  Back to cited text no. 1
Guo Y, Zhong Z, Yang C, Hu J, Jiang Y, Liang Z, et al. Epi-GTBN: An approach of epistasis mining based on genetic Tabu algorithm and Bayesian network. BMC Bioinformatics 2019;20:444.  Back to cited text no. 2
Visscher PM, Brown MA, McCarthy MI, Yang J. Five years of GWAS discovery. Am J Hum Genet 2012;90:7-24.  Back to cited text no. 3
VanderWeele TJ. Epistatic interactions. Stat Appl Genet Mol Biol 2010;9:Article 1.  Back to cited text no. 4
Bae SC. Epidemiology and etiology of rheumatoid arthritis. J Korean Med Assoc 2010;53:843-52.  Back to cited text no. 5
Julià A, Moore J, Miquel L, Alegre C, Barceló P, Ritchie M, et al. Identification of a two-loci epistatic interaction associated with susceptibility to rheumatoid arthritis through reverse engineering and multifactor dimensionality reduction. Genomics 2007;90:6-13.  Back to cited text no. 6
Meng Y, Yang Q, Cuenco KT, Cupples LA, Destefano AL, Lunetta KL. Two-stage approach for identifying single-nucleotide polymorphisms associated with rheumatoid arthritis using random forests and Bayesian networks. BMC Proc 2007;1 Suppl 1:S56.  Back to cited text no. 7
Ritchie MD, Bartlett J, Bush WS, Edwards TL, Motsinger AA, Torstenson ES. Exploring epistasis in candidate genes for rheumatoid arthritis. BMC Proc 2007;1 Suppl 1:S70.  Back to cited text no. 8
Carlton VE, Hu X, Chokkalingam AP, Schrodi SJ, Brandon R, Alexander HC, et al. PTPN22 genetic variation: Evidence for multiple variants associated with rheumatoid arthritis. Am J Hum Genet 2005;77:567-81.  Back to cited text no. 9
Plenge RM, Padyukov L, Remmers EF, Purcell S, Lee AT, Karlson EW, et al. Replication of putative candidate-gene associations with rheumatoid arthritis in >4,000 samples from North America and Sweden: Association of susceptibility with PTPN22, CTLA4, and PADI4. Am J Hum Genet 2005;77:1044-60.  Back to cited text no. 10
Yuan Z, Gao Q, He Y, Zhang X, Li F, Zhao J, et al. Detection for gene-gene co-association via kernel canonical correlation analysis. BMC Genet 2012;13:83.  Back to cited text no. 11
Clarke GM, Pettersson FH, Morris AP. A comparison of case-only designs for detecting genexgene interaction in rheumatoid arthritis using genome-wide case-control data in genetic analysis workshop 16. BMC Proc 2009;3 Suppl 7:S73.  Back to cited text no. 12
Buil A, Martinez-Perez A, Perera-Lluna A, Rib L, Caminal P, Soria JM. A new gene-based association test for genome-wide association studies. BMC Proc 2009;3 Suppl 7:S130.  Back to cited text no. 13
Greene CS, Penrod NM, Kiralis J, Moore JH. Spatially Uniform ReliefF (SURF) for computationally-efficient filtering of gene-gene interactions. BioData Min 2009;2:5.  Back to cited text no. 14
Yang C, He Z, Wan X, Yang Q, Xue H, Yu W. SNPHarvester: A filtering-based approach for detecting epistatic interactions in genome-wide association studies. Bioinformatics 2009;25:504-11.  Back to cited text no. 15
D'Angelo GM, Rao D, Gu CC. Combining least absolute shrinkage and selection operator (LASSO) and principal-components analysis for detection of gene-gene interactions in genome-wide association studies. BMC Proc 2009;3 Suppl 7:S62.  Back to cited text no. 16
Li J, Tang R, Biernacka JM, de Andrade M. Identification of gene-gene interaction using principal components. BMC Proc 2009;3 Suppl 7:S78.  Back to cited text no. 17
Liang X, Gao Y, Lam TK, Li Q, Falk C, Yang XR, et al. Identifying rheumatoid arthritis susceptibility genes using high-dimensional methods. BMC Proc 2009;3 Suppl 7:S79.  Back to cited text no. 18
Briggs FB, Ramsay PP, Madden E, Norris JM, Holers VM, Mikuls TR, et al. Supervised machine learning and logistic regression identifies novel epistatic risk factors with PTPN22 for rheumatoid arthritis. Genes Immun 2010;11:199-208.  Back to cited text no. 19
Peng Q, Zhao J, Xue F. A gene-based method for detecting gene-gene co-association in a case-control association study. Eur J Hum Genet 2010;18:582-7.  Back to cited text no. 20
Beretta L, Santaniello A, van Riel PL, Coenen MJ, Scorza R. Survival dimensionality reduction (SDR): Development and clinical application of an innovative approach to detect epistasis in presence of right-censored data. BMC Bioinformatics 2010;11:416.  Back to cited text no. 21
Yang C, Wan X, Yang Q, Xue H, Yu W. Identifying main effects and epistatic interactions from large-scale SNP data via adaptive group lasso. BMC Bioinformatics 2010;11 Suppl 1:S18.  Back to cited text no. 22
Wang Y, Liu X, Robbins K, Rekaya R. AntEpiSeeker: Detecting epistatic interactions for case-control studies using a two-stage ant colony optimization algorithm. BMC Res Notes 2010;3:117.  Back to cited text no. 23
Chanda P, Zhang A, Ramanathan M. Modeling of environmental and genetic interactions with AMBROSIA, an information-theoretic model synthesis method. Heredity (Edinb) 2011;107:320-7.  Back to cited text no. 24
Yoshida M, Koike A. SNPInterForest: A new method for detecting epistatic interactions. BMC Bioinformatics 2011;12:469.  Back to cited text no. 25
Dai H, Charnigo RJ, Becker ML, Leeder JS, Motsinger-Reif AA. Risk score modeling of multiple gene to gene interactions using aggregated-multifactor dimensionality reduction. BioData Min 2013;6:1.  Back to cited text no. 26
Walters RK, Laurin C, Lubke GH. Epi2Loc: An R package to investigate two-locus epistatic models. Twin Res Hum Genet 2014;17:272-8.  Back to cited text no. 27
Walters RK, Laurin C, Lubke GH. EpiPen: An r package to investigate two-locus epistatic models (Twin Research and Human Genetics). Twin Res Hum Genet 2014;17:1-6.  Back to cited text no. 28
Sengupta Chattopadhyay A, Hsiao CL, Chang CC, Lian IB, Fann CS. Summarizing techniques that combine three non-parametric scores to detect disease-associated 2-way SNP-SNP interactions. Gene 2014;533:304-12.  Back to cited text no. 29
Li J, Huang D, Guo M, Liu X, Wang C, Teng Z, et al. A gene-based information gain method for detecting gene-gene interactions in case-control studies. Eur J Hum Genet 2015;23:1566-72.  Back to cited text no. 30
Emily M. AGGrEGATOr: A Gene-based GEne-Gene interActTiOn test for case-control association studies. Stat Appl Genet Mol Biol 2016;15:151-71.  Back to cited text no. 31
Abo Alchamlat S, Farnir F. KNN-MDR: A learning approach for improving interactions mapping performances in genome wide association studies. BMC Bioinformatics 2017;18:184.  Back to cited text no. 32
Sinoquet C, Niel C. Enhancement of a Stochastic Markov Blanket Framework with ant Colony Optimization, to Uncover Epistasis in Genetic Association Studies. ESANN 2018 – Proceedings, European Symposium on Artificial Neural Networks, Computational Intelligenceand Machine Learning; 2018. p. 673-8.  Back to cited text no. 33
Guan B, Zhao Y, Li Y. DESeeker: Detecting epistatic interactions using a two-stage differential evolution algorithm. IEEE Access 2019;7:69604-13.  Back to cited text no. 34


  [Table 1], [Table 2], [Table 3]


Similar in PUBMED
   Search Pubmed for
   Search in Google Scholar for
 Related articles
Access Statistics
Email Alert *
Add to My List *
* Registration required (free)

  In this article
Epistasis Intera...
Review on Comput...
Performance Measures
Article Tables

 Article Access Statistics
    PDF Downloaded210    
    Comments [Add]    

Recommend this journal