Build Prognostic Models Using LASSO and Ridge Regression

Prepares data, splits it into training and testing sets, and fits LASSO and Ridge regression models for survival analysis. Evaluates model performance using cross-validation and optionally generates time-dependent ROC curves for visual assessment of predictive accuracy.

Usage

PrognosticModel(
  x,
  y,
  scale = FALSE,
  seed = 123456,
  train_ratio = 0.7,
  nfold = 10,
  plot = TRUE,
  palette = "jama",
  cols = NULL
)

Arguments

x

A matrix or data frame of predictor variables (features).

y

A data frame of survival outcomes with two columns: survival time and event status.

scale

Logical indicating whether to scale predictor variables. Default is `FALSE`.

seed

Integer seed for random number generation to ensure reproducibility. Default is `123456`.

train_ratio

Numeric proportion of data for training (e.g., 0.7). Default is `0.7`.

nfold

Integer number of folds for cross-validation. Default is `10`.

plot

Logical indicating whether to plot ROC curves. Default is `TRUE`.

palette

String specifying color palette. Default is `"jama"`.

cols

Optional vector of colors for ROC curves. If `NULL`, uses default palette.

Value

A list containing:

lasso_result

Results from LASSO model including coefficients and AUC

ridge_result

Results from Ridge model including coefficients and AUC

train.x

Training data with sample IDs

Author

Dongqiang Zeng

Examples

# \donttest{
if (requireNamespace("glmnet", quietly = TRUE) &&
  requireNamespace("survival", quietly = TRUE)) {
  library(survival)
  imvigor210_sig <- load_data("imvigor210_sig")
  imvigor210_pdata <- load_data("imvigor210_pdata")
  pdata_prog <- data.frame(
    ID = imvigor210_pdata$ID,
    OS_days = as.numeric(imvigor210_pdata$OS_days),
    OS_status = as.numeric(imvigor210_pdata$OS_status)
  )
  prognostic_result <- PrognosticModel(
    x = imvigor210_sig, y = pdata_prog,
    scale = TRUE, seed = 123456,
    train_ratio = 0.7, nfold = 10, plot = FALSE
  )
}
#> ℹ Processing data
#> ℹ Splitting data into training and validation sets
#> ℹ Running LASSO
#> ℹ Running RIDGE REGRESSION
#> ✔ Model fitting complete
# }