compute_all_tests {kernscr}

Testing pathway risk association

Description

This functions computes p-values frm score tests of genetic pathway risk association in 5 different models

Usage

compute_all_tests(
  data,
  ind_gene = 7:ncol(data),
  num_perts = 1000,
  Ws = NULL,
  rho = NA,
  kernel = c("linear", "gaussian", "poly"),
  d = 2,
  pca_thres = 0.9,
  get_ptb_pvals = FALSE,
  ...
)

Arguments

data

a data.frame of N rows and set up as the output from sim_SCR_data with columns:

• XR: time to recurrence / death / censoring

• XD: time to death / censoring

• DeltaR: Indicator of censoring (0), recurrence (1), or death (2) for this earliest time XR

• DeltaD: Indicator of censoring (0) or death (1)

• XPFS: time to recurrence / death / censoring (=XR)

• DeltaPFS: Indicator of censoring (0) or recurrence or death, whichever came first (1)

• Z_1,...,Z_P: genomic variables

ind_gene

columns indices of genes in the pathway of interest. Default is 7:ncol(data)).

num_perts

number of perturbations used. Default is 1000.

Ws

optional inputed perturbations, should be a vector of length N x num_perts containing i.i.d. realization of a random variable with mean=0 and variance=1.

rho

a vector of rhos, such as one found created from the range returned by findRhoInterval, used for tuning non-linear kernel. Only used if kernel is not "linear". Default is NA. Currently not available for use by user-defined kernels.

kernel

a character string indicating which kernel is used. Possible values (currently implemented) are "linear", "gaussian" or "poly". Otherwise, this can also be a user defined kernel function. See genericKernelEval.

d

if kernel is "poly", the polynomial power. Default is 2 (quadratic kernel).

pca_thres

a number between 0 and 1 giving the threshold to be used for PCA. Default is 0.9. If NULL, no PCA is performed.

get_ptb_pvals

a logical flag indicating whether perturbed p-values should be returned as part of the results. Default is FALSE.

...

extra parameters to be passed to a user-defined kernel.

Value

either a vector of p-values for 5 different models with names:

• "SCR": Semi-Competing Risks

• "PFS": Progression Free Survival

• "CR": Competing Risks

• "OS": Overall Survival

• "SCR_alt": SCR allowing different tuning parameters for the two event time processes

or else if get_ptb_pvals is TRUE, a list with 2 elements:

• "obs_pvals": a vector containing the observed p-values for each of the 5 models as described above

• "null_pvals_perts": a matrix of dimensions num_perts x 5 containing the corresponding perturbed p-values

References

Neykov M, Hejblum BP, Sinnot JA, Kernel Machine Score Test for Pathway Analysis in the Presence of Semi-Competing Risks, submitted, 2016.

Examples

## First generate some Data
feat_m_fun <- function(X){
 sin(X[,1]+X[,2]^2)-1
}
feat_d_fun <-  function(X){
 (X[,4]-X[,5])^2/8
}
mydata <- sim_SCR_data(data_size = 400, ncol_gene_mat = 20, feat_m = feat_m_fun,
                      feat_d = feat_d_fun, mu_cen = 40, cov=0.5)

#initial range
ind_gene <- c(7:ncol(mydata))
my_rho_init <- seq(0.01, 20, length=300)*length(ind_gene)
range(my_rho_init)

if(interactive()){
# compute the interval for rho
rho_set <- findRhoInterval(tZ=t(mydata[,ind_gene]), rho_init = my_rho_init, kernel="gaussian")
rho_set
range(my_rho_init) # good to check that the interval produced here is strictly contained in rho_init
# otherwise, expand rho.init and rerun

rhos <- exp(seq(log(rho_set[1]),log(rho_set[2]), length=50))

# run the tests with Gaussian kernel
compute_all_tests(data = mydata, num_perts=1000, rho=rhos, kernel="gaussian")
# run the tests with linear kernel
compute_all_tests(data=mydata, num_perts=1000, kernel="linear")
}