sangerCoxSurv: Fit Cox survival to all variants in a .vcf.gz file from...
In gwasurvivr: gwasurvivr: an R package for genome wide survival analysis

Description Usage Arguments Details Value Examples

View source: R/sangerCoxSurv.R

Performs survival analysis using Cox proportional hazard models on imputed genetic data stored in compressed VCF files.

sangerCoxSurv(vcf.file, covariate.file, id.column, sample.ids = NULL,
  time.to.event, event, covariates, inter.term = NULL,
  print.covs = "only", out.file, maf.filter = 0.05,
  info.filter = NULL, chunk.size = 5000, verbose = TRUE,
  clusterObj = NULL)

`vcf.file`	character(1) path to VCF file.
`covariate.file`	matrix(1) comprising phenotype (time, event) and additional covariate data.
`id.column`	character(1) providing exact match to sample ID column from `covariate.file`
`sample.ids`	character vector with sample ids to include in analysis
`time.to.event`	character(1) string that matches time column name in pheno.file
`event`	character(1) string that matches event column name in pheno.file
`covariates`	character vector with matching column names in pheno.file of covariates of interest
`inter.term`	character(1) string giving the column name of the covariate that will be added to the interaction term with SNP (e.g. `term*SNP`). See details.
`print.covs`	character(1) string of either `"only"`, `"all"` or `"some"`, defining which covariate statistics should be printed to the output. See details.
`out.file`	character(1) string with output name
`maf.filter`	integer(1) filter out minor allele frequency below threshold (i.e. 0.005 will filter MAF > 0.005)
`info.filter`	integer(1) of imputation quality score filter (i.e. 0.7 will filter info > 0.7)
`chunk.size`	integer(1) number of variants to process per thread
`verbose`	logical(1) for messages that describe which part of the analysis is currently being run
`clusterObj`	A cluster object that can be used with the `parApply` function. See details.

Testing for SNP-covariate interactions: User can define the column name of the covariate that will be included in the interaction term. For example, for given covariates a and b, where c is defined as the inter.term the model will be: ~ a + b + c + SNP + c*SNP.

Printing results of other covariates: print.covs argument controls the number of covariates will be printed as output. The function is set to only by default and will only print the SNP or if an interaction term is given, the results of the interaction term (e.g. SNP*covariate). Whereas, all will print results (coef, se.coef, p.value etc) of all covariates included in the model. some is only applicable if an interaction term is given and will print the results for SNP, covariate tested for interaction and the interaction term. User should be mindful about using the all option, as it will likely slow down the analysis and will increase the output file size.

User defined parallelization: This function uses parApply from parallel package to fit models to SNPs in parallel. User is not required to set any options for the parallelization. However, advanced users who wish to optimize it, can provide a cluster object generated by makeCluster family of functions that suits their need and platform.

Saves two text files directly to disk: .coxph extension containing CoxPH survival analysis results. .snps_removed extension containing SNPs that were removed due to low variance or user-defined thresholds.

vcf.file <- system.file(package="gwasurvivr",
                       "extdata",
                       "sanger.pbwt_reference_impute.vcf.gz")
pheno.fl <- system.file(package="gwasurvivr",
                        "extdata",
                     "simulated_pheno.txt")
pheno.file <- read.table(pheno.fl,
                         sep=" ",
                         header=TRUE,
                         stringsAsFactors = FALSE)
pheno.file$SexFemale <- ifelse(pheno.file$sex=="female", 1L, 0L)
sample.ids <- pheno.file[pheno.file$group=="experimental",]$ID_2
sangerCoxSurv(vcf.file=vcf.file,
              covariate.file=pheno.file,
              id.column="ID_2",
              sample.ids=sample.ids,
              time.to.event="time",
              event="event",
              covariates=c("age", "SexFemale", "DrugTxYes"),
              inter.term=NULL,
              print.covs="only",
              out.file="sanger_example",
              info.filter=0.3,
              maf.filter=0.005,
              chunk.size=50,
              verbose=TRUE,
              clusterObj=NULL)