Our aim of this study is to propose a methodological issue in detecting gene-gene interaction. We chose five commonly used methods and apply them to a schizophrenia data. Methods included traditional methods (chi-square test, LRM), Bayesian approach (BEAM), tree based model (CART), and combinatorial method (MDR). We also propose a haplotype-based study in gene-gene interaction. Using the haplotype based marker could give more information. If a haplotype block is highly associated with disease, the true disease gene (SNP) could be in the haplotype block. In the present study, we find that SNPs rsDAO_13 and rsDAO_7 have strong main effect.
SNPs rsDAO_6, rsDAO_7, and rsG72_16 have strong gene-gene interaction effects. It can give the biologist a suggestion to type more markers in these genes for future analysis.
In order to compare the predictive ability of these methods, we used cross-validation approach and defined a prediction rule. LRM shows the best predictive ability in our data.
Table 1. Marker’s Information
# Name Position ObsHET PredHET HWpval %Geno MAF Alleles 1 rsNRG1_6 32198397 0.348 0.381 0.0153 99.1 0.256 G:T 2 rsNRG1_14 32525521 0.301 0.322 0.0973 88.9 0.201 C:T 3 rsNRG1_8 32541620 0.092 0.096 0.3498 99.6 0.05 T:C 4 rsNRG1_1 32572900 0.336 0.334 0.9863 99.6 0.212 A:G 5 rsNRG1_13 32593784 0.492 0.494 0.9683 99.2 0.443 T:C 6 rsNRG1_11 32641669 0.467 0.482 0.3734 99.1 0.405 A:T 7 rsNRG1_2 32705627 0.127 0.133 0.3115 99.8 0.072 T:C
8 rsNRG1_E_1 32733529 0 0 1 89.5 0 G:G
9 rsCACNG2_3 35302102 0.425 0.427 0.9 99.6 0.31 G:T 10 rsCACNG2_23 35318530 0.459 0.478 0.2592 99 0.395 A:G 11 rsCACNG2_16 35351483 0.477 0.494 0.3322 99.4 0.447 A:G 12 rsCACNG2_15 35351741 0.495 0.497 0.9549 96.9 0.459 A:G 19 rsG72_11 103840146 0.061 0.059 0.8786 97.2 0.031 C:A 20 rsG72_1 104908896 0.443 0.456 0.4583 99.2 0.351 C:A 21 rsG72_2 104915349 0.447 0.456 0.5988 99.4 0.351 C:T 22 rsG72_E_1 104916613 0.141 0.15 0.1193 99.8 0.082 C:T 23 rsG72_16 104927525 0.325 0.341 0.2251 89.2 0.218 G:C
24 rsG72_E_4 104927538 0 0 1 89.4 0 A:A
25 rsG72_17 104927721 0.338 0.347 0.4586 98.7 0.223 A:T 26 rsG72_6 104940236 0.241 0.253 0.196 99.8 0.149 C:T 27 rsG72_7 104940237 0.031 0.03 1 99.3 0.015 G:A 28 rsG72_E_3 104940243 0.004 0.004 1 89.5 0.002 C:T 29 rsG72_13 104941175 0.47 0.475 0.7798 99.9 0.388 C:A 30 rsG72_14 104941217 0.045 0.044 1 96.5 0.023 A:T 31 rsDAO_2 107797548 0.117 0.114 0.761 96.4 0.061 G:A 32 rsDAO_3 107797907 0.009 0.009 1 99.3 0.005 G:C 33 rsDAO_5 107798175 0.097 0.096 1 99 0.051 G:A
Table 1. Marker’s Information (Cont’d)
34 rsDAO_6 107801621 0.477 0.474 0.9449 95.5 0.387 C:A 35 rsDAO_7 107801849 0.479 0.47 0.646 93.7 0.378 G:A 36 rsDAO_8 107801872 0.483 0.47 0.4581 98.8 0.377 T:G 37 rsDAO_E_1 107803071 0.006 0.006 1 89.4 0.003 C:A 38 rsDAO_9 107805607 0.123 0.12 0.5754 99.4 0.064 G:C 39 rsDAO_10 107807701 0.124 0.121 0.5991 96 0.064 T:G
40 rsDAO_E_2 107808165 0 0 1 89.5 0 T:T
41 rsDAO_11 107811039 0.123 0.119 0.6063 98.2 0.064 G:A 42 rsDAO_13 107816559 0.231 0.225 0.5161 99.8 0.129 C:T 43 rsDISC1_24 229829230 0.29 0.305 0.1719 99.6 0.188 G:A 44 rsDISC1_40 229829627 0.454 0.471 0.3114 100 0.38 A:G 45 rsDISC1_E_1 229896474 0.026 0.026 1 99.7 0.013 C:T 46 rsDISC1_E_3 229896886 0.001 0.001 1 89.4 0.001 C:T 47 rsDISC1_E_4 229897110 0.021 0.021 1 99.9 0.011 C:A 48 rsDISC1_27 229925804 0.472 0.481 0.6406 99.8 0.402 G:A 49 rsDISC1_16 229926137 0.212 0.211 1 97 0.12 G:A 50 rsDISC1_2 229961231 0.493 0.5 0.7202 93.6 0.493 G:A 51 rsDISC1_35 229969633 0.342 0.35 0.5493 99.9 0.226 C:T 52 rsDISC1_E_5 229973212 0.296 0.296 1 99.6 0.181 C:T 53 rsDISC1_E_6 229973396 0.077 0.084 0.046 99.8 0.044 G:C 54 rsDISC1_3 229997671 0.189 0.202 0.0964 99.4 0.114 T:C 55 rsDISC1_4 230020768 0.05 0.051 0.9073 98.9 0.026 G:A 56 rsDISC1_12 230024766 0.465 0.465 1 99.4 0.368 G:A
57 rsDISC1_34 230068001 0 0 1 99.2 0 A:A
58 rsDISC1_26 230069015 0.466 0.468 0.9427 99.9 0.374 A:G 59 rsDISC1_5 230143129 0.014 0.014 1 98.4 0.007 A:T 60 rsDISC1_E_7 230211221 0.214 0.207 0.4108 99.6 0.117 A:T 61 rsDISC1_38 230228487 0.134 0.129 0.3688 99.6 0.069 G:T 62 rsDISC1_20 230240183 0.364 0.381 0.2247 99.1 0.256 G:T 63 rsDISC1_36 230241611 0.253 0.266 0.1683 99.2 0.158 A:G 64 rsDISC1_7 230242818 0.201 0.207 0.4383 99.7 0.117 G:T 65 rsDISC1_15 230243610 0.421 0.442 0.1649 99.8 0.33 C:T
Table 2.a. Single marker effects detected by the five methods in genotype-based data
rank Chisq LRM BEAM CART MDR
1 rsDAO_13 rsDAO_13 rsDAO_13 rsDAO_7 rsDAO_7
2 rsDAO_7 rsDAO_7 rsDAO_7 rsDAO_6
3 rsDAO_6 rsDAO_6 rsNRG1_6 rsNRG1_6
4 rsNRG1_6 rsNRG1_6 rsCACNG2_3 rsDAO_13
5 rsDISC1_38 rsDISC1_38 rsDISC1_38 rsDAO_8
Table 2.b. Single marker effects detected by the four methods in haplotype-based data
rank Chisq LRM BEAM MDR
1 DAO_block1 DAO_block1 DAO_block1 DAO_block1
2 G72_block2 G72_block2 CACNG2_block2 rsNRG1_6
3 rsNRG1_6 rsNRG1_6 DISC1_block4
4 CACNG2_block2 CACNG2_block2 DISC1_block2
5 rsDISC1_38 rsDISC1_38 G72_block2
Table 3.a. Two-way interaction detected by the five methods in genotype-based data
rank Chisq LRM BEAM CART MDR
1 rsDAO_6 rsDAO_7 rsDAO_6 rsDAO_7 rsDISC1_E_7 rsDISC1_E_4 rsDAO_7 rsDAO_8 rsNRG1_14 rsG72_16
2 rsNRG1_6 rsDAO_6 rsDAO_7 rsDAO_8 rsDAO_6 rsDAO_7 rsNRG1_6 rsDAO_6
3 rsNRG1_6 rsDAO_7 rsDAO_6 rsDAO_8 rsDISC1_3 rsDAO_7
4 rsDAO_7 rsDAO_13 rsDISC1_20 rsNRG1_6 rsDISC1_16 rsNRG1_6
5 rsDAO_6 rsDAO_13 rsDISC1_16 rsDISC1_20 rsDAO_6 rsDAO_7
Table 3.b. Two-way interaction detected by the four methods in haplotype-based data
rank Chisq LRM BEAM MDR
1 rsNRG1_6 G72_block2 rsDISC1_E_7 G72_block2 No two-way interaction detected DISC1_block3 DAO_block1 2 DAO_block1 G72_block2 rsNRG1_6 CACNG2_block2 DISC1_block1 DAO_block1
3 G72_block2 CACNG2_block2 rsDISC1_E_7 rsCACNG2_3 DAO_block1 G72_block1
4 rsNRG1_6 DAO_block1 G72_block2 CACNG2_block2 DISC1_block4 DAO_block1 5 rsNRG1_6 CACNG2_block2 rsDISC1_38 CACNG2_block2 DISC1_block5 DAO_block1
Table 4.a. Three-way interaction detected by the five methods in genotype-based data
rank Chisq BEAM MDR Table 4.b. Three-way interaction detected by the three methods in haplotype-based data
Table 5. Summary of rsDAO_13
Haplotype Case Control Total Odds Odds ratio CI AAGGTGC
Table 7. Summary of DAO_block1 (Cont’d)
Haplotype Case Control Total Odds Odds ratio CI CGTCGAC
Table 7. Summary of DAO_block1 (Cont’d)
Haplotype Case Control Total Odds Odds ratio CI CGTGTGT
Table 8. Summary of rsDAO_6*rsDAO_7
Genotype Case Control Total Odds Odds ratio CI
AA*AA 65 47 112 1.3830 1.4784 (0.9537 , 2.2916)
CC*GG 145 155 300 0.9355 Reference Total 513 376 889
Table 9. Average prediction error across 100 CVs
Chisq LRM BEAM CART MDR
one-way 0.471283784 0.476047297 0.471148649 0.486824324 0.473783784 two-way 0.464207618 0.448881209 0.488123798 0.477674915 0.470942832
three-way 0.495776846 0.491696159 0.494607021
Figure 3.a. Haplotype block in DISC1
Figure 3.b. Haplotype block in NRG1
Figure 3.c. Haplotype block in DAO
Figure 3.d. Haplotype block in G72
Figure 3.e. Haplotype block in CACNG2
Figure 4. Box-plot of prediction error of one-way interaction
Figure 5. Box-plot of prediction error of two-way interaction
Figure 6. Box-plot of prediction error of three-way interaction
References
1. Briollais L, Wang Y, Rajendram I, Onay V, Shi E, Knight J, Ozcelik H:
Methodological issues in detecting gene-gene interactions in breast cancer susceptibility: a population-based study in Ontario. BMC Medicine 2007, 5:22.
2. Heidema AG, Boer JM, Nagelkerke N, Mariman EC, A DLvd, Feskens EJ:
The challenge for genetic epidemiologists: how to analyze large numbers of SNPs in relation to complex diseases. BMC Genetics 2006, 7:23
3. Hwu H-G, Faraone SV, Chih-Min Liu WJC, Liu S-K, Shieh M-H, Hwang T-J, Tsuang M-M, OuYang W-C, Chen C-Y, Chen C-C et al: Taiwan Schizophrenia Linkage Study: The Field Study. American Journal of Medical Genetics Part B 2005, 134B:30-36.
4. Faraone SV, Hwu H-G, Liu C-M, Chen WJ, Tsuang M-M, Liu S-K, Shieh M-H, Hwang T-J, Ou-Yang W-C, Chen C-Y et al: Genome Scan of Han Chinese Schizophrenia Families From Taiwan: Confirmation of Linkage to 10q22.3. Am J Psychiatry 2006, 163:1760-1766.
5. Iachine I: Basics of human genetics. In Statistical Methods in Genetic Epidemiology; 2004.
6. Balding DJ: A tutorial on statistical methods for population association studies. NATURE REVIEWS GENETICS 2006, 7:781-791.
7. Zhang Y, Liu JS: Bayesian inference of epistatic interactions in case-control studies. NATURE GENETICS 2007, 39:1167-1173.
8. Clark LA, Pregibon D: Tree-based models. In Statistical Models in S Edited by: Chambers JM, Hastie TJ. Pacific Grove, California: Wadsworth
and Brooks/Cole Advanced Books and Software; 1992:377-419.
9. Roger J. Lewis, M.D.: An Introduction to Classification and Regression Tree (CART) Analysis. http://www.saem.org/download/lewis1.pdf
10. Hahn LW, Ritchie MD, Moore JH: Multifactor dimensionality reduction software for detecting gene-gene and gene-environment interactions.
BIOINFORMATICS 2003, 19:376-382.
11. Barrett JC, Fry B, Maller J, Daly MJ: Haploview: analysis and visualization of LD and haplotype maps. BIOINFORMATICS APPLICATIONS NOTE 2005, 21:263-265.
12. Stephens M, Smith NJ, Donnelly P: A New Statistical Method for Haplotype Reconstruction from Population Data. Am J Hum Genet 2001, 68:978-989.
13. Stephens M, Donnelly P: A comparison of bayesian methods for haplotype reconstruction. American Journal of Human Genetics 2003, 73:1162-1169.
14. Thornton-Wells TA, Moore JH, Haines JL: Genetics, statistics and human disease: analytical retooling for complexity. TRENDS in Genetics 2004, 20:640-647.