RCASPAR (version 1.18.0)

RCASPAR-package: A package for survival time prediction based on a piecewise baseline hazard Cox regression model.

Description

The package is the R-version of the C-based software CASPAR (Kaderali,2006). It is meant to help predict survival times in the presence of high-dimensional explanatory co-variates. The model is a piecewise baseline hazard Cox regression model with an Lq-norm based prior that selects for the most important regression coefficients, and in turn the most relevant co-variates for survival analysis. It was primarily tried on gene expression and aCGH data, but can be used on any other type of high-dimensional data and in disciplines other than biology and medicine.

Details

Package:

RCASPAR

Type:

Package

Version:

1.0

Date:

2010-08-23

License: GPL(>=3) LazyLoad:

yes

References

The basic model is based on the Cox regression model as first introduced by Sir David Cox in: Cox,D.(1972).Regression models & life tables. Journal of the Royal Society of Statistics, 34(2), 187-220. The extension of the Cox model to its stepwise form was adapted from: Ibrahim, J.G, Chen, M.-H. & Sinha, D. (2005). Bayesian Survival Analysis (second ed.). NY: Springer. as well as Kaderali, Lars.(2006) A Heirarchial Bayesian Approach to Regression and its Application to Predicting Survival Times in Cancer Patients. Aachen: Shaker The prior on the regression coefficients was adopted from: Mazur, J., Ritter,D.,Reinelt, G. & Kaderali, L. (2009). Reconstructing Non-Linear dynamic Models of Gene Regulation using Stochastic Sampling. BMC Bioinformatics, 10(448).

Examples

## Eg.(1): A simple example performed with a training and validation set:
data(Bergamaschi)
data(survData)
  ## Generate prediction:
result <- STpredictor_BLH(geDataS=Bergamaschi[1:27, 1:2], survDataS=survData[1:27, 9:10], geDataT=Bergamaschi[28:82, 1:2], survDataT=survData[28:82, 9:10], q = 1, s = 1, a = 1.558, b = 0.179
, cut.off=15, groups = 3, method = "CG", noprior = 1, extras = list(reltol=1))
 ## Plot a KM plot with both long and short survivors:
kmplt_svrl(long=result$long_survivors, short=result$short_survivors,title="KM plot of long and short survivors")
  ## Determine the area under the curve of AUROC curves vs. time to see the performance of the predictor given the chosen parameters and the current partitioning into training
  ## and validation sets:
survivAURC(Stime=result$predicted_STs$True_STs,status=result$predicted_STs$censored, marker=result$predicted_STs$Predicted_STs, time.max=20)
 ## Perform a log-rank test to see if the difference between the long and short survivors is significant:
logrnk(dataL=result$long_survivors, dataS=result$short_survivors)

## Eg.(2): A simple example performed with cross validation:
data(Bergamaschi)
data(survData)
  ## Generate prediction:
STpredictor_xvBLH(geData=Bergamaschi[1:40,1:2], survData=survData[1:40,9:10], k = 10, cut.off = 10, q = 1, s = 1, a = 1.558, b = 0.179, groups = 3, method = "BFGS", noprior = 1, extras = list(reltol=1))
 ## Plot a KM plot with both long and short survivors:
kmplt_svrl(long=result$long_survivors, short=result$short_survivors,title="KM plot of long and short survivors")
  ## Determine the area under the curve of AUROC curves vs. time to see the performance of the predictor given the chosen parameters and the current partitioning into training
  ## and validation sets:
survivAURC(Stime=result$predicted_STs$True_STs,status=result$predicted_STs$censored, marker=result$predicted_STs$Predicted_STs, time.max=20)
 ## Perform a log-rank test to see if the difference between the long and short survivors is significant:
logrnk(dataL=result$long_survivors, dataS=result$short_survivors)