SNPを用いた2x3テストに関するメモ

  • 関係するRのパッケージはこちら
  • This is something on the two types of test for 2x3 tables.
  • When we do case-control association test for a SNP, we get a 2x3 table.
  • For a 2x3 table from an autosomal SNP, we cat perform
    • (1) 2x3 genotypic test with df 2,
    • (2) dominant test for which we make 2x2 table from the 2x3 table according to the genotypes
    • (3) also we can do recessive test
    • (4) then we can do additive model test (trend test)
    • (5) somebody may make 2x2 table from 2x3 by counting the number of chromosomes with one allele and the other. In this case the total of 4 cells of 2x2 table is 2N where N is the sum of individuals.
  • (4) and (5) are both testing allele's effect on the phenotype in the additive model but we should do (4) not (5).
  • This can be clear when you apply (4) and (5) to X choromosomal SNPs in male population
  • When we genotype males for X chromosomal SNPs, all males have homozygous genotypes. Therefore when we give the 2x3 table as
x,y,z
a,b,c

y=b=0.
  • When we perform trend test(4), essentially we perform 2x2 test for a table
x,z
a,c
  • But when we make 2x2 table from this by counting number of chromosomes,
2x,2z
2z,2c
  • is the table we get.
  • Total sample size gets doubled, so chi-square value will be also doubled.
  • Actually when autosomal SNPs are in HWE perfectly, trend test and 2x2 of -chromosomal number test are mutually identical...
  • You can double-check this by using the excel file (here).
  • Input
100,200,90
100,200,110
  • in the light blue area.
  • Then trend test result is 1.00063 (cell B20) and allele comparison of chromosomal number is 1.0006(cell B 17).
  • But when you input
100,200,190
100,200,210
  • their values are 0.45733 and 0.5335.
  • As in X chromosome for males,
100,0,200
100,0,240
  • then their values are 1.14082 and 2.2816 = 1.14082*2