Calculates a threshold to identify a significant QTL (M.P. Boer & J.T.N.M. Thissen).
Options
PRINT = string token |
What to print (summary); default summ |
|---|---|
POPULATIONTYPE = string token |
Type of population (BC1, DH1, F2, RIL, BCxSy, CP); must be set |
THRMETHOD = string token |
Which method to use (bonferroni, liji); default liji |
STATISTICTYPE = string token |
Which type of test statistic to use (wald, minlog10p); default minl |
ALPHALEVEL = scalar |
Defines the genome-wide significance level; default 0.05 |
DISTANCE = scalar |
Distance between evaluation points for THRMETHOD=bonferroni; default 4 |
DF = scalar |
Degrees of freedom for the Wald test; default 1 |
Parameters
CHROMOSOMES = factors |
Chromosome for each locus; must be set |
|---|---|
POSITIONS = variates |
Position on the chromosome for each locus; must be set |
ADDITIVEPREDICTORS = pointers |
The additive genetic predictors |
ADD2PREDICTORS = pointers |
The second (paternal) additive genetic predictors if POPULATIONTYPE is CP |
DOMINANCEPREDICTORS = pointers |
The dominance genetic predictors if POPULATIONTYPE is F2 or CP |
THRESHOLD = scalars |
Saves the calculated threshold |
Description
QTHRESHOLD calculates a genome wide significance threshold to use as a critical value to reject the null hypothesis of no QTL effect. The genome-wide type I error rate is defined by the option ALPHALEVEL. The threshold is based on a modified Bonferroni correction. The THRMETHOD option specifies the method for calculating the number of tests to used as the denominator. The default setting, liji, uses the effective number of independent tests, as described by Li & Ji (2005). Alternatively, the setting bonferroni assumes one independent test at every fixed distance on the genome, defined by the DISTANCE option (default 4 centiMorgans). By default, the threshold is expressed as the P value on a -log10 scale, but you can set option STATISTICTYPE=Wald to use the absolute Wald test statistic instead. Marker and map information must be supplied by the ADDITIVEPREDICTORS, CHROMOSOMES and POSITIONS parameters. The DOMINANCEPREDICTORS parameter can supply dominance genetic predictors for population types F2, RIL, BCxSy and CP, and the ADD2PREDICTORS parameter can supply the second (paternal) additive genetic predictors for population type CP. The corresponding degrees of freedom for the Wald test must be set by the DF parameter; this is equal to 1 in a single-environment QTL analysis, or to the number of environments in a multi-environment QTL analysis.
The calculated threshold can be saved using the THRESHOLD parameter. By default the threshold is printed, but you can suppress this by setting option PRINT=*.
Options: PRINT, POPULATIONTYPE, THRMETHOD, STATISTICTYPE, ALPHALEVEL, DISTANCE, DF.
Parameters: CHROMOSOMES, POSITIONS, ADDITIVEPREDICTORS, ADD2PREDICTORS, DOMINANCEPREDICTORS, THRESHOLD.
Method
QTHRESHOLD calculates a genome-wide significance threshold based on a modified Bonferonni correction, where the effective number of tests is used as the denominator instead of the total number of tests. By default the procedure estimates the effective number of independent tests by a singular value decomposition of the correlation matrix between all markers (see Li & Ji 2005 or Cheverud 2001). Alternatively, QTHRESHOLD assumes that the effective number of tests along the genome can be approximated by n independent tests:
n = ceiling(L/D)
where L is the total genome length (in cM), D is the distance between evaluation points (in cM) supplied by the DISTANCE option, and ceiling(x) gives the smallest integer not less than x. If the DISTANCE option is unset, n is set to the length of the CHROMOSOMES variate.
Using the Bonferonni correction, the genome wide significance threshold T is approximated by Χ2df(1-α/n), where df is the number of degrees of freedom.
Action with RESTRICT
Restrictions are not allowed.
References
Cheverud, J.M. (2001). A simple correction for multiple comparisons in interval mapping genome scans. Heredity, 87, 52-58.
Li, J, & Ji, L. (2005). Adjusting multiple testing in multilocus analyses using the eigenvalues of a correlation matrix. Heredity, 95, 221-227.
See also
Procedures: FDRBONFERRONI, FDRMIXTURE.
Commands for: Statistical genetics and QTL estimation.
Example
CAPTION 'QTHRESHOLD example'; STYLE=meta
SPLOAD [PRINT=*] '%GENDIR%/Examples/F2maizemarkers.gwb'; SHEETNAME='LOCI'
& '%GENDIR%/Examples/F2maizemarkers.gwb'; SHEETNAME='ADDPREDICTORS'
QTHRESHOLD [THRMETHOD=liji; STATISTICTYPE=minlog; ALPHA=0.05]\
CHROMOSOMES=mkchr; POSITIONS=mkpos; ADDITIVEPREDICTORS=addpred
QTHRESHOLD [THRMETHOD=liji; STATISTICTYPE=wald; ALPHA=0.05; DF=1]\
CHROMOSOMES=mkchr; POSITIONS=mkpos; ADDITIVEPREDICTORS=addpred
QTHRESHOLD [THRMETHOD=liji; STATISTICTYPE=wald; ALPHA=0.05; DF=10]\
CHROMOSOMES=mkchr; POSITIONS=mkpos; ADDITIVEPREDICTORS=addpred
QTHRESHOLD [THRMETHOD=bonferroni; STATISTICTYPE=wald; DF=1; DISTANCE=*]\
CHROMOSOMES=mkchr; POSITIONS=mkpos
QTHRESHOLD [THRMETHOD=bonferroni; STATISTICTYPE=wald; DF=1; DISTANCE=50]\
CHROMOSOMES=mkchr; POSITIONS=mkpos
QTHRESHOLD [THRMETHOD=bonferroni; STATISTICTYPE=minlog; DISTANCE=*]\
CHROMOSOMES=mkchr; POSITIONS=mkpos