You are here
- Home
- Archive
- Volume 43, Issue 1
- Analysis of RUNX1 binding site and RAPTOR polymorphisms in psoriasis: no evidence for association despite adequate power and evidence for linkage
Responses
PDF
Other responses
Jump to comment:
- James T. ElderPublished on: 27 April 2016
- Anne M. BowcockPublished on: 27 April 2016
- Published on: 27 April 2016Authors' response to Bowcock et al.
- James T. Elder, physician-scientist
- Other Contributors:
- Philip E. Stuart, Rajan P. Nair, Gonçalo R. Abecasis, Ioana Nistor, Ravi Hiremagalore, Nicholas V. Chia Zhaohui S. Qin, Rachel A. Thompson, Stefan Jenisch, Michael Weichenthal, Jennifer Janiga, Henry W. Lim, Enno Christophers,and John J.Voorhees
Show MoreDear Editor,
In response to the comments of Helms et al.: We recognize that in the presence of substantial heterogeneity, much larger sample sizes may be required to replicate an effect. Thus, we cannot rule out the possibility that the variants in the RUNX binding site and the RAPTOR gene make a substantial contribution to psoriasis susceptibility in some settings. Nevertheless, in our sample, which includes...
Show LessDear Editor,
In response to the comments of Helms et al.: We recognize that in the presence of substantial heterogeneity, much larger sample sizes may be required to replicate an effect. Thus, we cannot rule out the possibility that the variants in the RUNX binding site and the RAPTOR gene make a substantial contribution to psoriasis susceptibility in some settings. Nevertheless, in our sample, which includes 2,261 genotyped individuals and 1,143 psoriatics appropriate for FBAT association analysis, we see no significant evidence (p<.05) of association between psoriasis and these SNPs. To our knowledge, this is the largest sample yet examined for these markers.
Individuals interested in a more detailed point-by-point response are invited to go to the following URL: http://www.psoriasis.umich.edu/jmgresponse.pdf.
Philip Stuart1
Rajan P. Nair1
Gonçalo R. Abecasis2
Ioana Nistor1
Ravi Hiremagalore1
Nicholas V. Chia1
Zhaohui S. Qin2
Rachel A. Thompson1
Stefan Jenisch3
Michael Weichenthal4
Jennifer Janiga5
Henry W. Lim5
Enno Christophers4
John J.Voorhees1
James T. Elder1,6,7*
Departments of 1Dermatology and 6Radiation Oncology (Cancer Biology), University of Michigan Medical School, Ann Arbor, MI
2Center for Statistical Genetics, Department of Biostatistics, School of Public Health, University of Michigan, Ann Arbor, MI
Departments of 3Immunology and 4Dermatology, University of Kiel, Kiel, Germany
5Department of Dermatology, Henry Ford Hospital, Detroit, MI
7Department of Dermatology, Ann Arbor VA Health System, Ann Arbor, MI
*correspondence should be addressed to J.T.E. (jelder@umich.edu)
Conflict of Interest:
None declared. - Published on: 27 April 2016Analysis of RUNX1 binding site and RAPTOR polymorphisms in psoriasis
- Anne M. Bowcock, Professor
- Other Contributors:
- Cynthia Helms, Li Cao, James G. Krueger, Ellen M .Wijsman, Derek Gordon, Michael Heffernan, Jason Robarge1, Jurg Ott, Pui Y. Kwok, Alan Menter
Show MoreDear Editor,
In the recently published report "Analysis of RUNX1 Binding Site and RAPTOR Polymorphisms in Psoriasis: No Evidence for Association Despite Adequate Power and Evidence for Linkage"[1] the authors report failure to see association of psoriasis susceptibility to a region of chromosome 17q25. This region was first identified with linkage analysis by our group[2], and we recently reported evidence for a...
Show LessDear Editor,
In the recently published report "Analysis of RUNX1 Binding Site and RAPTOR Polymorphisms in Psoriasis: No Evidence for Association Despite Adequate Power and Evidence for Linkage"[1] the authors report failure to see association of psoriasis susceptibility to a region of chromosome 17q25. This region was first identified with linkage analysis by our group[2], and we recently reported evidence for association to two unlinked loci from this chromosomal region.[3] In this instance one associated SNP allele leads to loss of a putative RUNX binding site. It is important to address the approach and findings of Stuart et al. (2005)[1] since they omit important factors that may be required for replicating associations between genetic loci and complex traits:
1) The statement about adequate power in the title, and the discussion in the text is misleading. The power these authors report is only accurate for the parameters they considered. They fail to consider the inherent heterogeneity of a disease such as psoriasis and do not incorporate a number of critical parameters into their power analyses. These include: (i) locus heterogeneity, (ii) disequilibrium (r2) less than 1.0, (iii) phenotype and/or genotype misclassification and (iv) the effects and interactions of environmental risk factors with genetic factors (currently unknown for complex traits). All are realistic factors and all can substantially reduce power to detect either linkage and/or association.[4-15]
2) The issue of genetic heterogeneity is of great importance in complex human disease genetics. It has been demonstrated that sample size needed to replicate a positive linkage study exceeds, by a considerable margin, the original sample size, particularly when factors such as heterogeneity are involved.[16] Effect sizes in published positive studies tend to be inflated over what one gets after much more data collection.[17] Such studies are simply more fortunate with respect to the component of the study that leads to publication, but it does not mean that they are false positive results. Since Stuart et al. (2005) did not include heterogeneity into their power calculations they cannot have addressed this issue.
3) Genotype misclassification seems particularly relevant in this work. The authors comment that they reconstructed haplotypes using the methods implemented in Merlin and PHASE (with the exception of FBAT). This reconstruction was done to address issues about inflation in type I error due to SNP markers that are not in linkage equilibrium. However, it has been well documented that haplotype reconstruction methods such as these are prone to misclassification18, and that misclassification probabilities may be large for nuclear families like those studied in this work.[19]
4) It is not clear why the authors did not perform a TDT analysis using all the affected individuals in the pedigrees, i.e. perform a linkage analysis with the tightly linked markers? The authors report that they had 1,285 affected individuals in total. Yet, they chose to perform an association analysis with TDT, therefore using a maximum of 351 trios (Table 1), or less than 1/3 of the available data. The TDT linkage analysis using all affecteds was the analysis performed in our study[3] and so without a corresponding analysis on their part, a direct comparison is not possible. It is also interesting that although the ratios of transmitted to untransmitted alleles are by no means significant in the study of Stuart et al. (2005) there is a trend in all pedigrees in the direction we report: (for SNP9/rs734232 the ratio of the associated A allele in transmitted versus non-transmitted chromosomes is 235:219).
5) Another issue that could be relevant is the pedigree ascertainment. Although the ratio of sampled affected to unaffected individuals in the pedigrees in our study[3] does not appear to be very different from the same ratio in the pedigrees of Stuart et al. (2005)[1], there could also be an effect if there is a difference in the way the pedigrees come to attention in the two studies. It is possible that the complete pedigrees (including unsampled individuals) are larger with more affected individuals in the former study[3] than in the latter.[1] The two samples could then end up with different fractions of a particular risk haplotype. A comparison of haplotype frequencies obtained by Stuart et al. (2005)[1] and Helms et al. (2003)[3] does indeed suggest some systematic difference between the data sets and/or analyses. Stuart et al. (2005)[1] state that the combined frequency for the two most common haplotypes was between .992-.999 for the two regions (depending somewhat on methods of reconstruction). In the study from Helms et al. (2003)[3] , the combined frequency of the two most common haplotypes on non-transmitted chromosomes was 0.94-0.95. This suggests either that the reconstructions are flawed, or that the samples differ. The difference could either lead to difference in power, since if one has more different haplotypes there may be more informative transmissions. Alternatively it could indicate a fundamental sampling difference.
6) Given that we report five SNPs driving association in our families, it is surprising that Stuart et al. (2005)[1] did not attempt to examine the haplotypes determined by more than just three associated SNPs. As we report, in a case/control study, evidence for association with associated haplotypes demarcated by these five driving SNPs was greater than with single SNPs.[3]
7) Stuart et al. (2005)[1] also state that "only one marker in this peak exceeded the p=0.05 level of significance after the most stringent level of correction for multiple testing", making independent confirmation critical. However, our level of correction for multiple testing employed the FDR method[20,21], because employing a Bonferroni correction for association studies with multiple SNPs is considered excessively conservative, generally obliterating any signal in large-scale associations studies. With the FDR method, nine markers in the SLC9A3R1/NAT9 remained significant. There are several other studies reporting that adjusting for multiple testing in studies is less important than other concerns.[22,23]
8) A final general comment is warranted. We believe that chromosome 17q25 may harbor a number of loci for psoriasis susceptibility. The issue of clustering of genes for complex traits has been reported by others[24,25], and is thought to facilitate identification of susceptibility loci with linkage analyses. Stuart et al. (2005) also see linkage of psoriasis to 17q25, in a region harboring the putative RUNX site downstream from SLC9A3R1/ NAT9 and in a region distinct from that seen in two independent studies of large multiplex families[2,26], but where we also see linkage in our large set of nuclear families.[3] Previous studies with simulation data suggest that single major gene analysis of complex traits has good power to localize genes to a specific chromosome, but power to localize beyond the chromosomal level may be significantly compromised.[27] A full understanding of the contribution of different susceptibility factors from this region will be necessary as we seek to understand the genetic complexity of this disease.
C Helms1
L Cao1
JG Krueger2
EM Wijsman3
F Chamian2
D Gordon4
M Heffernan5
JA Wright-Daw1
J Robarge1
J Ott4
P-Y Kwok6
A Menter7
AM Bowcock11Department of Genetics
Washington University School of Medicine
St. Louis
Missouri, 63110, USA.2Laboratory for Investigative Dermatology
The Rockefeller University
New York
NY 10021, USA.3Div. of Medical Genetics and Dept. Biostatistics
University of Washington
Seattle
Washington, 98195, USA.4Lab of Statistical Genetics
The Rockefeller University
New York
NY 10021, USA.5Div. Of Dermatology
Washington University School of Medicine
St. Louis
Missouri, 6310, USA.6Department of Dermatology
Cardiovascular Research Institute, and Center for Human Genetics
University of California
San Francisco
CA 94143, USA.7Department of Internal Medicine
Division of Dermatology
Baylor University Medical Center
Dallas
Texas, 75246 USA.*Correspondence should be addressed to A.M.B. (email: bowcock@genetics.wustl.edu).
References
1. Stuart P, Nair RP, Abecasis GR, Nistor I, Hiremagalore R, Chia NV, Qin ZS, Thompson RA, Jenisch S, Weichenthal M, Janiga J, Lim HW, Christophers E, Voorhees JJ, Elder JT. Analysis of RUNX1 binding site and RAPTOR polymorphisms in psoriasis: No evidence for Association Despite Adequate power and Evidence for linkage. J Med Genet 2005 (online publication).
2. Tomfohrde J, Silverman A, Barnes R, Fernandez-Vina MA, Young M, Lory D, Morris L, Wuepper KD, Stastny P, Menter, A. Gene for familial psoriasis susceptibility mapped to the distal end of human chromosome 17q. Science 1994;264:1141-5.
3. Helms C, Cao L, Krueger JG, Wijsman EM, Chamian F, Gordon D, Heffernan M, Daw J, Robarge J, Ott J, Kwok P-Y, Menter A, Bowcock AM. A putative RUNX1 binding site variant between SLC9A3R1 and NAT9 is associated with susceptibility to psoriasis. Nat Genet 2003;35:349-56.
4. Sillanpaa MJ, Auranen K. Replication in genetic studies of complex traits. Ann Hum Genet 2004;68:646-57.
5. Bross I. Misclassification in 2 x 2 tables. Biometrics 1954;10:478-86.
6. Mote VL, Anderson RL. An investigation of the effect of misclassification on the properties of chisquare-tests in the analysis of categorical data. Biometrika 1965;52:95-109.
7. Cochran WG. Errors of measurement in statistics. Technometrics 1968;10:637-666.
8. Ott J. Analysis of human genetic linkage: Johns Hopkins, Baltimore, 1999.
9. Gordon D, Matise TC, Heath SC, Ott J. Power loss for multiallelic transmission/disequilibrium test when errors introduced: GAW11 simulated data. Genet Epidemiol Suppl 1999;17:S587-S92.
10. Gordon D, Heath SC, Liu X, Ott J. A transmission/ disequilibrium test that allows for genotyping errors in the analysis of single-nucleotide polymorphism data. Am J Hum Genet 2001;69:371-80.
11. Gordon D, Finch SJ, Nothnagel M, Ott J. Power and sample size calculations for case-control genetic association tests when errors are present: application to single nucleotide polymorphisms. Hum Hered 2002;54:22-33.
12. Gordon D, Haynes C, Johnnidis C, Patel SB, Bowcock AM, Ott J. A transmission disequilibrium test for general pedigrees that is robust to the presence of random genotyping errors and any number of untyped parents. Eur J Hum Genet 2004;12:752-61.
13. Zou G, Zhao H. The impacts of errors in individual genotyping and DNA pooling on association studies. Genet Epidemiol 2004;26:1-10.
14. Zheng G, Tian X. The impact of diagnostic error on testing genetic association in case-control studies. Stat Med 2005;24:869- 82.
15. Edwards BJ, Haynes C, Levenstien MA, Finch SJ, Gordon D. Power and sample size calculations in the presence of phenotype errors for case/control genetic association studies. BMC Genet 2005;6:18.
16. Suarez BK, Hampe CL, van Eerdewegh P. Problems of replicating linkage claims in psychiatry. Washington: American Psychiatric Press, 1994.
17 Goring HHH, Terwilliger JD, Blangero J (2001) Large upward bias in estimation of locus-specific effects from genomewide scans Am J Hum Genet 69:1357-1369.
18. Niu T. Algorithms for inferring haplotypes. Genet Epidemiol 2004;27:334-347.
19. Lindholm E, Zhang J, Hodge SE, Greenberg DA. The reliability of haplotyping inference in nuclear families: misassignment rates for SNPs and microsatellites. Hum Hered 2004;57:117-127.
20. Benjamini Y, Yekutieli D. Quantitative Trait Loci Analysis using the False Discovery Rate. Genetics 2005.
21. Benjamini Y, Drai D, Elmer G, Kafkafi N, Golani I. Controlling the false discovery rate in behavior genetics research. Behav Brain Res 2001;125:279-84.
22. Perneger TV. Adjusting for multiple testing in studies is less important than other concerns. BMJ 1999;318:1288.
23. Witte JS, Elston RC, Schork NJ. Genetic dissection of complex traits. Nat Genet 1996;12:355-6.
24. Morel L, Blenman KR, Croker BP, Wakeland EK. The major murine systemic lupus erythematosus susceptibility locus, Sle1, is a cluster of functionally related genes. Proc Natl Acad Sci, USA 2001;98:1787-92.
25. Adeniji OA, Mrug MM, DiPalma JA. Not one but two inflammatory bowel disease susceptibility loci map to chromosome 16. Am J Gastroenterol 2002;97:2464-5.
26. Hwu WL, Yang CF, Fann CS, Chen CL, Tsai TF, Chien YH, Chiang SC, Chen CH, Hung SI, Wu JY, Chen YT. Mapping of psoriasis to 17q terminus. J Med Genet 2005;42:152-8.
27. Gordon, D., Hoh, J., Finch, S.J., Levenstien, M.A., Edington, J., Li, W., Majewski, J. and Ott, J. Two approaches for consolidating results from genome scans of complex traits: selection methods and scan statistics. Genet Epidemiol 2001;21 Suppl 1, S396-402.
Conflict of Interest:
None declared.