Coding-variant Allelic Series Test

Data

To run an allelic series test, there are 4 key inputs:

A numeric annotation vector, of the same length as the number of variants, coded as 1 for benign missense variants (BMVs), 2 for damaging missense variants (DMVs), and 3 for protein truncating variants (PTVs).
A covariates matrix, with as many rows as subjects and including columns such as age and sex. If omitted, defaults to an intercept only. If provided, an intercept will automatically be added if not detected.
A genotype matrix, with subjects as rows and variants as columns. The number of columns should correspond to the length of the annotation vector.
A numeric phenotype vector, either continuous or binary.

The example data used below were generated using the DGP function provided with the package. The data set includes 100 subjects, 300 variants, and a continuous phenotype. The true effect sizes follow an allelic series, with magnitudes proportional to c(1, 2, 3) for BMVs, DMVs, and PTVs respectively.

set.seed(101)
n <- 100
data <- AllelicSeries::DGP(
  n = n,
  snps = 300,
  beta = c(1, 2, 3) / sqrt(n),
)

# Annotations.
anno <- data$anno
head(anno)

## rs1 rs2 rs3 rs4 rs5 rs6 
##   1   1   2   2   1   2

# Covariates.
covar <- data$covar
head(covar)

##      intercept         age sex          pc1        pc2         pc3
## [1,]         1 -0.77465210   1  0.539531581  1.2134116  0.63478280
## [2,]         1 -0.03492052   0  0.009270313  1.5256400 -0.62027221
## [3,]         1 -0.20782643   1 -0.777242512  1.2319750 -0.73311857
## [4,]         1  0.69939529   0 -0.357743663  0.1747664 -0.29754242
## [5,]         1  0.47420055   1 -1.083115262 -1.3005348 -0.68338808
## [6,]         1 -1.15038152   0 -0.539141451 -0.5159589  0.01971706

# Genotypes.
geno <- data$geno
head(geno[,1:5])

##      rs1 rs2 rs3 rs4 rs5
## [1,]   0   0   0   0   0
## [2,]   0   0   0   0   0
## [3,]   0   0   0   0   0
## [4,]   0   0   0   0   0
## [5,]   0   0   0   0   0
## [6,]   0   1   0   0   0

# Phenotype.
pheno <- data$pheno
head(pheno)

## [1] 0.2958143 1.7294107 0.4886318 1.0201730 1.0304183 1.9704552

The example data generated by the preceding are available under vignettes/vignette_data.

Running the alleic series test

The COding-variant Allelic Series Test (COAST) is run using the COAST function. By default, the output of COAST includes a data.frame of counts showing the number of alleles, variants, and carriers in each class that contributed to the test, and a data.frame of p-values, with the omni test denoting the final, overall p-value. Inspection of the component p-values may be useful for determining which model(s) detected evidence of an association. In the present case, the baseline count model provided the greatest power.

results <- AllelicSeries::COAST(
  anno = anno,
  geno = geno,
  pheno = pheno,
  covar = covar
)
show(results)

## Effect Sizes:
##         test beta    se
## 1       base 0.05 0.068
## 2       base 0.06 0.079
## 3       base 0.59 0.175
## 4        ind 0.31 0.256
## 5        ind 0.22 0.216
## 6        ind 0.66 0.218
## 7  max_count 0.06 0.048
## 8    max_ind 0.44 0.130
## 9  sum_count 0.07 0.031
## 10   sum_ind 0.19 0.058
## 
## 
## Counts:
##   anno alleles variants carriers
## 1    1     186      164       84
## 2    2     127      110       74
## 3    3      27       26       23
## 
## 
## P-values:
##           test   type     pval
## 1     baseline burden 6.62e-03
## 2          ind burden 9.11e-03
## 3    max_count burden 2.32e-01
## 4      max_ind burden 8.15e-04
## 5    sum_count burden 1.48e-02
## 6      sum_ind burden 1.09e-03
## 7 allelic_skat   skat 6.83e-02
## 8         omni   omni 4.68e-03

The effect sizes data.frame is accessed via:

results@Betas

##         test       beta         se
## 1       base 0.05309214 0.06836091
## 2       base 0.06427266 0.07917711
## 3       base 0.58612733 0.17472435
## 4        ind 0.30597919 0.25595872
## 5        ind 0.21943023 0.21607933
## 6        ind 0.66475646 0.21829831
## 7  max_count 0.05719340 0.04783470
## 8    max_ind 0.43655519 0.13041078
## 9  sum_count 0.07437087 0.03052869
## 10   sum_ind 0.19038906 0.05828998

the counts data.frame via:

results@Counts

##   anno alleles variants carriers
## 1    1     186      164       84
## 2    2     127      110       74
## 3    3      27       26       23

and the p-values data.frame via:

results@Pvals

##           test   type        pval
## 1     baseline burden 0.006618953
## 2          ind burden 0.009107121
## 3    max_count burden 0.231834498
## 4      max_ind burden 0.000815325
## 5    sum_count burden 0.014846685
## 6      sum_ind burden 0.001089858
## 7 allelic_skat   skat 0.068349669
## 8         omni   omni 0.004683243

Ultra-rare variants

Effect size estimation for ultra-rare variants is challenging due to the scarcity of carriers on which to base the estimate. Following works such as SAIGE-GENE+, we recommend collapsing variants with a minor allele count (MAC) \(\leq 10\) into an aggregate variant, separately within each annotation category. The CollapseGeno utility is provided for this purpose. The min_mac specifies the threshold for retention as an individual variant (e.g. min_mac = 11 will collapse variants with a MAC \(\leq 10\)). The output is a list containing the annotation vector anno and genotype matrix geno after collapsing. In addition, a mapping vars is provided showing which variants were collapsed to create each aggregate variant. Note that:

CollapseGeno will change the order of the variants in the genotype matrix.
No aggregate variant will be created within annotation categories where no variants have a MAC below the threshold.
Collapsing ultra-rare variants does not guarantee that the resulting aggregate variant will itself have a MAC \(\geq\) the threshold. To ensure that only variants with a minimum MAC enter the association test, a min_mac threshold should also be supplied to COAST.

set.seed(102)

# Generate data.
n <- 1e3
data <- AllelicSeries::DGP(
  n = n,
  snps = 10,
  prop_anno = c(1, 1, 1) / 3
)

# Collapse ultra-rare variants.
collapsed <- AllelicSeries::CollapseGeno(
  anno = data$anno, 
  geno = data$geno,
  min_mac = 11
)

# Variants collapsed to form each aggregate variant.
head(collapsed$vars)

##   anno            vars
## 1    1     rs1;rs4;rs5
## 2    2             rs8
## 3    3 rs2;rs3;rs7;rs9

# Run COAST on the collapsed data.
results <- AllelicSeries::COAST(
  anno = collapsed$anno,
  covar = data$covar,
  geno = collapsed$geno,
  pheno = data$pheno,
  min_mac = 10
)

Different numbers of annotation categories

COAST was originally intended to operate on the benign missense variants, damaging missense variants, and protein truncating variants within a gene, but it has been generalized to allow for an arbitrary number of discrete annotation categories. The following example simulates and analyzes data with 4 annotation categories. The main difference when analyzing a different number of annotation categories is that the weight vector should be specified, and should have length equal to the number of possible annotation categories. COAST will run, albeit with a warning, if there are possible annotation categories to which no variants are assigned (e.g. a gene contains no PTVs).

set.seed(102)

# Generate data.
n <- 1e2
data <- AllelicSeries::DGP(
  n = n,
  snps = 400,
  beta = c(1, 2, 3, 4) / sqrt(n),
  prop_anno = c(0.4, 0.3, 0.2, 0.1),
  weights = c(1, 1, 1, 1)
)

# Run COAST-SS.
results <- AllelicSeries::COAST(
  anno = data$anno,
  covar = data$covar,
  geno = data$geno,
  pheno = data$pheno,
  weights = c(1, 2, 3, 4)
)
show(results)

## Effect Sizes:
##         test  beta    se
## 1       base  0.01 0.065
## 2       base  0.23 0.062
## 3       base  0.19 0.096
## 4       base  0.61 0.116
## 5        ind -0.10 0.253
## 6        ind  0.53 0.205
## 7        ind  0.31 0.169
## 8        ind  0.96 0.186
## 9  max_count  0.18 0.035
## 10   max_ind  0.47 0.090
## 11 sum_count  0.11 0.018
## 12   sum_ind  0.19 0.035
## 
## 
## Counts:
##   anno alleles variants carriers
## 1    1     194      169       86
## 2    2     159      131       78
## 3    3      71       61       51
## 4    4      41       39       31
## 
## 
## P-values:
##           test   type     pval
## 1     baseline burden 9.67e-10
## 2          ind burden 5.86e-07
## 3    max_count burden 4.26e-07
## 4      max_ind burden 1.65e-07
## 5    sum_count burden 5.37e-10
## 6      sum_ind burden 8.24e-08
## 7 allelic_skat   skat 3.27e-06
## 8         omni   omni 4.11e-09

Test options

apply_int = TRUE applies the rank-based inverse normal transformation from the RNOmni package. This transformation is expected to improve power for phenotypes that have a skewed or kurtotic (e.g. long-tailed) distribution. It is applied by default in the case of continuous phenotype, and is ignored in the case of a binary phenotype.

AllelicSeries::COAST(
  anno = anno,
  geno = geno,
  pheno = pheno,
  covar = covar,
  apply_int = TRUE
)

include_orig_skato_all = TRUE includes standard SKAT-O applied to all variants as a component of the omnibus test, while include_orig_skato_ptv = TRUE includes standard SKAT-O applied to PTVs only. In cases where other annotation schemes are used, the annotation for the PTV class can be specified via ptv_anno. Including standard SKAT-O as a component of the omnibus test can improve power to detect associations between the phenotype and genes that may not be allelic series.

AllelicSeries::COAST(
  anno = anno,
  geno = geno,
  pheno = pheno,
  covar = covar,
  include_orig_skato_all = TRUE,
  include_orig_skato_ptv = TRUE,
  ptv_anno = 3
)

is_pheno_binary = TRUE is required to indicate that the supplied phenotype is binary, and should be analyzed using a logistic regression model.

AllelicSeries::COAST(
  anno = anno,
  geno = geno,
  pheno = 1 * (pheno > 0),
  covar = covar,
  is_pheno_binary = TRUE
)

min_mac is used to filter the variant set to those containing a minimum minor allele count (MAC). The following example filters to only those variants with a MAC of at least 2:

AllelicSeries::COAST(
  anno = anno,
  geno = geno,
  pheno = pheno,
  covar = covar,
  min_mac = 2
)

return_omni_only = TRUE is used to return p_omni only when the component p-values are not of interest:

AllelicSeries::COAST(
  anno = anno,
  geno = geno,
  pheno = pheno,
  covar = covar,
  return_omni_only = TRUE
)

score_test = TRUE specifies the use of a score-type allelic series burden test. The default of score_test = FALSE specifies a Wald-type allelic series burden test.

AllelicSeries::COAST(
  anno = anno,
  geno = geno,
  pheno = pheno,
  covar = covar,
  score_test = TRUE
)

weights specifies the relative phenotypic effects of BMVs, DMVs, and PTVs. An increasing pattern such as the default setting of weights = c(1, 2, 3) targets allelic series. Setting weights = c(1, 1, 1) would target a genetic architecture where all variants have equivalent expected magnitudes.

AllelicSeries::COAST(
  anno = anno,
  geno = geno,
  pheno = pheno,
  covar = covar,
  weights = c(1, 2, 3)
)