Association of Vascular Endothelial Growth Factor (VEGF) polymorphisms with
childhood asthma and airway remodeling phenotypes
Online Depository
Methods for derivation of study-wide significance threshold:
To control type I error, the primary analysis of association of was adjusted for
multiple comparisons by adopting a significance threshold adjusted for the number of
independent tests performed. Because VEGF polymorphisms demonstrate very strong LD
(see Results, Figure 1A, and Supplemental Table S1), adjustment for the number of typed
markers (10) would be overly conservative and thus inappropriate. Therefore, we first
calculated the effective number of independent marker loci tested (MeffLi) using the
method of spectral decomposition (SpD) of matrices of pairwise LD proposed by
Nyholt19 as modified by Li and Ji20, as implemented in SNPSpD
(http://gump.qimr.edu.au/general/daleN/ SNPSpD/). This approach defines MeffLi by (i)
calculating the correlation matrix (i.e. pairwise LD) across all markers using all available
genotype data; (ii) measuring the collective correlation across the set of markers as the
variance of the eigenvalues from this LD matrix; and (iii) using this measure of collective
correlation to calculate the proportional reduction in the number of independent markers.
In the CAMP population, the calculated collective correlation was high (variance of the
observed eigenvalues = 1.3023) resulting in an estimated MeffLi of 8 markers, and an
adjusted alpha of 0.0064 (see Supplemental Table S1). We note that because our primary
hypothesis testing evaluated three phenotypes under two genetic models, we further
adjusted for six additional tests per marker, resulting in an experiment-wise significance
threshold of <0.001.
Table S1: Single Nucleotide Polymorphism Spectral Decomposition (SNPSpD)
-----------------------------------------------------------------------Matrix of pairwise LD correlations for all markers:
1
2
3
4
5
6
7
8
9
10
1
2
1
-0.35
-0.43
-0.30
-0.56
-0.28
-0.01
-0.02
0.01
-0.00
-0.35
1
0.77
0.33
0.56
0.29
0.04
-0.00
-0.07
-0.22
3
4
5
6
7
8
9
10
-0.43 -0.30 -0.56 -0.28 -0.01 -0.02 0.01 -0.00
0.77 0.33 0.56 0.29 0.04 -0.00 -0.07 -0.22
1
0.66 0.72 0.64 -0.02 -0.04 0.03 -0.05
0.66 1
0.48 0.95 0.06 0.07 0.05 0.03
0.72 0.48 1
0.48 -0.11 -0.14 0.11 0.08
0.64 0.95 0.48 1
0.03 0.06 0.04 0.06
-0.02 0.06 -0.11 0.03 1
0.89 0.12 -0.07
-0.04 0.07 -0.14 0.06 0.89 1
0.14 -0.06
0.03 0.05 0.11 0.04 0.12 0.14 1
0.55
-0.05 0.03 0.08 0.06 -0.07 -0.06 0.55 1
Matrix derived using all genotyped subjects. Calculated Eigenvalues are 3.6918, 1.9559,
1.4764, 1.1012, 0.7320, 0.4552, 0.3206, 0.1308, 0.0936, and 0.0425, with a resultant
observed eigenvalue variance of 1.3023. The resultant number of effective independent
loci, based on the method of Li and Ji20 is 8, requiring an experiment-wide significance
threshold of 0.0064 to maintain the type I error rate at 0.05. To maintain an experimentwide error rate of 0.05, considering three phenotypes under two genetic models, and pvalue < 0.001 is required (0.0064/ [3 (phenotypes) x 2 (genetic models)]).
Online Data Supplement References:
19.
Nyholt DR. A Simple Correction for Multiple Testing for Single-Nucleotide
Polymorphisms in Linkage Disequilibrium with Each Other. The American
Journal of Human Genetics 2004; 74:765-9.
20.
Li J, Ji L. Adjusting multiple testing in multilocus analyses using the eigenvalues
of a correlation matrix. Heredity 2005; 95:221-7.