Prediction of sample arrays with the SAGA model.

saga_predict uses a Support Vector Machine (SVM) with a radial kernel to classify user samples either as transforming or untransforming. The SVM model is built with the integrated SAGA training data set and a toplist of relevant probes which are able to differentiate assay data into the mentioned risk states. The SVM is further optimized on the SAGA training data. Use at your own risk!

saga_predict(
  smplpath,
  matrix.train,
  labels.train,
  matrix.unknown,
  pData.Test,
  writeFile = 0,
  showRoc = 1
)

Arguments

smplpath	path to sample data
matrix.train	normalized, probe-averaged and batch-corrected SAGA training data.
labels.train	class labels (factors) for SAGA training data. Can either be "transforming" or "untransforming".
matrix.unknown	matrix of sample data with array names as row names and probes as column names.
pData.Test	SAGA sample or test data matrix
writeFile	default=1 for writing results to the sample folder
showRoc	default=1 for showing the ROC curve of the model performance; alternative: 0 for showing naught

Value

predictions Data frame with three columns. Column one shows the sample names, the second column shows the decision values of the svm function and column thress shows the predictions for the query assays either as transforming or untransforming.