Filters loci that show significant departure from Hardy-Weinberg Equilibrium
This function filters out loci showing significant departure from H-W proportions based on observed frequencies of reference homozygotes, heterozygotes and alternate homozygotes.
Loci are filtered out if they show HWE departure either in any one population (n.pop.threshold =1) or in at least X number of populations (n.pop.threshold > 1).
gl.filter.hwe( x, subset = "each", n.pop.threshold = 1, method_sig = "Exact", multi_comp = FALSE, multi_comp_method = "BY", alpha_val = 0.05, pvalue_type = "midp", cc_val = 0.5, min_sample_size = 5, verbose = NULL )
Name of the genlight object containing the SNP data [required].
Way to group individuals to perform H-W tests. Either a vector with population names, 'each', 'all' (see details) [default 'each'].
The minimum number of populations where the same locus has to be out of H-W proportions to be removed [default 1].
Method for determining statistical significance: 'ChiSquare' or 'Exact' [default 'Exact'].
Whether to adjust p-values for multiple comparisons [default FALSE].
Method to adjust p-values for multiple comparisons: 'holm', 'hochberg', 'hommel', 'bonferroni', 'BH', 'BY', 'fdr' (see details) [default 'fdr'].
Level of significance for testing [default 0.05].
Type of p-value to be used in the Exact method. Either 'dost','selome','midp' (see details) [default 'midp'].
The continuity correction applied to the ChiSquare test [default 0.5].
Minimum number of individuals per population in which perform H-W tests [default 5].
Verbosity: 0, silent or fatal errors; 1, begin and end; 2, progress log; 3, progress and results summary; 5, full report [default 2, unless specified using gl.set.verbosity].
There are several factors that can cause deviations from Hardy-Weinberg proportions including: mutation, finite population size, selection, population structure, age structure, assortative mating, sex linkage, nonrandom sampling and genotyping errors. Therefore, testing for Hardy-Weinberg proportions should be a process that involves a careful evaluation of the results, a good place to start is Waples (2015).
Note that tests for H-W proportions are only valid if there is no population substructure (assuming random mating) and have sufficient power only when there is sufficient sample size (n individuals > 15).
Populations can be defined in three ways:
Merging all populations in the dataset using subset = 'all'.
Within each population separately using: subset = 'each'.
Within selected populations using for example: subset = c('pop1','pop2').
Two different statistical methods to test for deviations from Hardy Weinberg proportions:
The classical chi-square test (method_sig='ChiSquare') based on the function
HWChisqof the R package HardyWeinberg. By default a continuity correction is applied (cc_val=0.5). The continuity correction can be turned off (by specifying cc_val=0), for example in cases of extreme allele frequencies in which the continuity correction can lead to excessive type 1 error rates.
The exact test (method_sig='Exact') based on the exact calculations contained in the function
HWExactStatsof the R package HardyWeinberg, and described in Wigginton et al. (2005). The exact test is recommended in most cases (Wigginton et al., 2005). Three different methods to estimate p-values (pvalue_type) in the Exact test can be used:
'dost' p-value is computed as twice the tail area of a one-sided test.
'selome' p-value is computed as the sum of the probabilities of all samples less or equally likely as the current sample.
'midp', p-value is computed as half the probability of the current sample + the probabilities of all samples that are more extreme.
The standard exact p-value is overly conservative, in particular for small minor allele frequencies. The mid p-value ameliorates this problem by bringing the rejection rate closer to the nominal level, at the price of occasionally exceeding the nominal level (Graffelman & Moreno, 2013).
Correction for multiple tests can be applied using the following methods
based on the function
'holm' is also known as the sequential Bonferroni technique (Rice, 1989). This method has a greater statistical power than the standard Bonferroni test, however this method becomes very stringent when many tests are performed and many real deviations from the null hypothesis can go undetected (Waples, 2015).
'hochberg' based on Hochberg, 1988.
'hommel' based on Hommel, 1988. This method is more powerful than Hochberg's, but the difference is usually small.
'bonferroni' in which p-values are multiplied by the number of tests. This method is very stringent and therefore has reduced power to detect multiple departures from the null hypothesis.
'BH' based on Benjamini & Hochberg, 1995.
'BY' based on Benjamini & Yekutieli, 2001.
The first four methods are designed to give strong control of the family-wise error rate. The last two methods control the false discovery rate (FDR), the expected proportion of false discoveries among the rejected hypotheses. The false discovery rate is a less stringent condition than the family-wise error rate, so these methods are more powerful than the others, especially when number of tests is large. The number of tests on which the adjustment for multiple comparisons is the number of populations times the number of loci.
gl.filter.hwe takes the argument
n.pop.threshold > 1 loci will be removed only if they are
significant (after adjustment if applied) out of hwe in >=
n.pop.threshold > 1.
Benjamini, Y., and Yekutieli, D. (2001). The control of the false discovery rate in multiple testing under dependency. Annals of Statistics, 29, 1165–1188.
Graffelman, J. (2015). Exploring Diallelic Genetic Markers: The Hardy Weinberg Package. Journal of Statistical Software 64:1-23.
Graffelman, J. & Morales-Camarena, J. (2008). Graphical tests for Hardy-Weinberg equilibrium based on the ternary plot. Human Heredity 65:77-84.
Graffelman, J., & Moreno, V. (2013). The mid p-value in exact tests for Hardy-Weinberg equilibrium. Statistical applications in genetics and molecular biology, 12(4), 433-448.
Hochberg, Y. (1988). A sharper Bonferroni procedure for multiple tests of significance. Biometrika, 75, 800–803.
Hommel, G. (1988). A stagewise rejective multiple test procedure based on a modified Bonferroni test. Biometrika, 75, 383–386.
Rice, W. R. (1989). Analyzing tables of statistical tests. Evolution, 43(1), 223-225.
Waples, R. S. (2015). Testing for Hardy–Weinberg proportions: have we lost the plot?. Journal of heredity, 106(1), 1-19.
Wigginton, J.E., Cutler, D.J., & Abecasis, G.R. (2005). A Note on Exact Tests of Hardy-Weinberg Equilibrium. American Journal of Human Genetics 76:887-893.
Other filter functions:
Custodian: Luis Mijangos -- Post to https://groups.google.com/d/forum/dartr
result <- gl.filter.hwe(x = bandicoot.gl) #> Starting gl.filter.hwe #> Processing genlight object with SNP data #> Analysing each population separately #> Loci examined: 1000 #> Deleted 128 loci with significant departure from HWE at alpha = 0.05 applied locus by locus #> Loci retained: 872 #> #> Adjustment of p-values for multiple comparisons vary with #> sample size #> Completed: gl.filter.hwe #> #>