mvstest {signatureSurvival}

Multivariate Cox proportional hazard survival analyses with multiple genes

Description

Function mvstest performs MVKMresult by iteration from the first gene to the last gene. The output result of musvtest is used to build a vector of weights for screening genes as prognostic signature or doing signature survival analysis.

Usage

mvstest(sdata,X,stn,gn, status,time,quant=c("No",-0.2,0.2))

Arguments

sdata

survival data containing p patients in row, survival status for death or recurrence or relapse, survival time such as day, month, or year, and covariates age, sex and/or smoking, tumor stage, or chemotherapy, and a set of genes in column.

X

string specifying one or multiple covariates such as age, sex, smoking, or tumor stage or genotype. The number of covariates in X is limited to 10.

stn

character value specifying gene symbol or gene id existed in sdata or int value specifying column number for the first gene in survival data.

gn

character value specifying gene symbol or gene id existed in sdata or int value specifying column number for the last gene in survival data.

status

string for survival status which must be binary variable: 1 for event occurrence and 0 for no event occurrence. status may be "death", "relapse" or "recurrence", depending on clinical data.

time

string for survival time and may be one of c("day", "month", "year"), depending on the clinical data. The default value is "month".

quant

vector for quantile, low and high values. the low and high values are used to define or classify low and high expression groups. If quantile is "yes" or "YES", then the low and high are non-negative precent values, for example, quant=c("yes",0.25,0.75). If quantile is "no", then the low and high values are z-scores, the low value may be negative. For example, quant=c("no",-0.2, 0.2). The quantile = "yes" or = "No" may produce different results of survival analysis. User should carefully choose quantile or no quantile according to the data. The default values are c("no",-0.2,0.2).

Details

Patient survival status is binary variable with 1 for event (such as death) and 0 for no event (such as alive). Genes have expression values(numeric values), which are used to calculate z-scores for classifying patients into two groups: high-expression patients and low-expression patients. SKMCresult performs univariate Cox proportional hazard survival analyses of patients with expression values of a specified gene and outputs hazard risk (HR), z-score and p-value of this specified gene. At first, user can run this function by performing msvtest to screen genes for prognostic signature by using HRs, z-scores, and p-values. Once getting a set of genes for signature, user can perform this function to build a weight vector using

w_i=\frac{log_{10}(p_i)}{\sum_{i=1}^g{log_{10}(p_i)}}

where p_i is p-value for Ward-test of gene i. For a patient, the signature score or expression value is given by weighting expression values of g genes in the signature:

y_j=\sum_{i}^g{w_ix_{ij}}

where x_{ij} is expression of gene i in patient j.

Value

output a matrix with n rows for n genes, m+1 columns for Hazard risk, hazard rate, standard error, z-value, p-value of gene and m covariates.

Note

All inputting parameters are not sensitive to upper or lower. That is, user can input upper or lower string or letter. For example, both time ="MONTH" or time = "month" work.

Author(s)

Yuan-De Tan
[e-mail]
[e-mail]
Yuguang Ban
[e-mail]

See Also

MVKMresult

Examples

data(GSE50081)

res<-mvstest(sdata=GSE50081,X=c("t.stage","n.stage"),stn=3550,gn=3555,
status="status",time="month",quant=c("No",-0.2,0.2))