isb.splsdrcox function

Iterative single-block sPLS-DRCOX Dynamic

This function performs a single-block sparse partial least squares deviance residual Cox analysis (sPLS-DRCOX) using the optimal components and variables identified in a previous cross-validation process. It builds the final model based on the selected hyperparameters.

isb.splsdrcox(
  X,
  Y,
  cv.isb,
  x.center = TRUE,
  x.scale = FALSE,
  remove_near_zero_variance = TRUE,
  remove_zero_variance = TRUE,
  toKeep.zv = NULL,
  remove_non_significant = FALSE,
  alpha = 0.05,
  MIN_EPV = 5,
  returnData = TRUE,
  verbose = FALSE
)

Arguments

• X: List of numeric matrices or data.frames. Explanatory variables. If the variables are qualitative, they must be transformed into binary variables.
• Y: Numeric matrix or data.frame. Response variables with two columns: "time" and "event". Accepted values for the event column are 0/1 or FALSE/TRUE for censored and event observations, respectively.
• cv.isb: Instance of class "Coxmos" and model "cv.iSB.sPLS-DRCOX-Dynamic". Used to retrieve the optimal components and variables for the sPLS Cox model.
• x.center: Logical. If TRUE, the X matrix is centered to zero means (default: TRUE).
• x.scale: Logical. If TRUE, the X matrix is scaled to unit variance (default: FALSE).
• remove_near_zero_variance: Logical. If TRUE, near-zero variance variables are removed (default: TRUE).
• remove_zero_variance: Logical. If TRUE, zero-variance variables are removed (default: TRUE).
• toKeep.zv: Character vector. Names of variables in X to retain despite near-zero variance filtering (default: NULL).
• remove_non_significant: Logical. If TRUE, non-significant variables/components in the final Cox model are removed through forward selection (default: FALSE).
• alpha: Numeric. Significance threshold (default: 0.05).
• MIN_EPV: Numeric. Minimum number of Events Per Variable (EPV) for the final Cox model. Limits the number of variables/components allowed (default: 5).
• returnData: Logical. If TRUE, returns the original and normalized X and Y matrices (default: TRUE).
• verbose: Logical. If TRUE, extra messages will be displayed (default: FALSE).

Returns

An object of class "Coxmos" and model "isb.splsdrcox_dynamic", containing:

• X: List with normalized X data:

  • data: Normalized X matrix (or NA if not returned).
  • x.mean: Mean values of the X matrix.
  • x.sd: Standard deviations of the X matrix.

• Y: List with normalized Y data:

  • data: Normalized Y matrix.
  • y.mean: Mean values of the Y matrix.
  • y.sd: Standard deviations of the Y matrix.

• survival_model: Fitted survival model (Cox proportional hazards model).

• list_spls_models: List of sPLS models computed for each block.

• n.comp: Number of components selected.

• n.varX: Number of variables selected per block.

• call: Function call.

• X_input: Original X matrix (or NA if not returned).

• Y_input: Original Y matrix (or NA if not returned).

• alpha: Significance threshold used.

• nsv: Variables removed due to non-significance.

• nzv: Variables removed due to near-zero variance.

• nz_coeffvar: Variables removed due to near-zero coefficient of variation.

• class: Model class.

• time: Time taken to run the analysis.

Details

The isb.splsdrcox_dynamic function fits a single-block sPLS-DRCOX model using the input data and the optimal components and variables determined from cross-validation. The function allows for centering and scaling of the data, and it offers the option to remove variables with near-zero variance, zero variance, or those that are non-significant based on a specified alpha level.

This method is particularly suited for high-dimensional data where there are many more variables than observations. The function can handle multiple blocks of data, and integrates them into a single model for Cox proportional hazards analysis.

Examples

data("X_multiomic")
data("Y_multiomic")
set.seed(123)
index_train <- caret::createDataPartition(Y_multiomic$event, p = .25, list = FALSE, times = 1)
X_train <- X_multiomic
X_train$mirna <- X_train$mirna[index_train,1:15]
X_train$proteomic <- X_train$proteomic[index_train,1:15]
Y_train <- Y_multiomic[index_train,]
vector <- list()
vector$mirna <- c(10, 15)
vector$proteomic <- c(10, 15)
cv <- cv.isb.splsdrcox(X_train, Y_train, max.ncomp = 1, vector = vector, n_run = 1, k_folds = 2)
model <- isb.splsdrcox(X_train, Y_train, cv)