Skip to main content

Table 1 Hyperparameter spaces used for tuning

From: Empowering individual trait prediction using interactions for precision medicine

Algorithm Hyperparameter Description Values
glmnet alpha Elastic net mixing parameter. alpha = 1 is the LASSO, alpha = 0 is the ridge penalty \(\left\{\mathrm{0,0.25,0.5,0.75,1}\right\}\)
ranger num.trees Number of trees 1000
  mtry Number of variables to possibly split at in each node \(\left[\mathrm{1,100}\right]\subset {\mathbb{N}}\)
  min.node.size Minimal node size \(\left[\mathrm{10,100}\right]\subset {\mathbb{N}}\)
MBMDRC min.cell.size Minimum number of samples with a specific genotype combination to be statistically relevant. If less, a cell is automatically labelled as \(O\) \(\left[\mathrm{0,50}\right]\subset {\mathbb{N}}\)
  alpha Significance level used to determine \(H\), \(L\) and \(O\) label of a cell \(\left(\mathrm{0.01,1}\right)\subset {\mathbb{R}}\)
  adjustment Adjustment for lower order marginal effects {NONE, CODOMINANT}
  order Number of SNPs to be considered in MDR models \(\left\{\mathrm{1,2}\right\}\)
  order.range Use order as upper limit? {TRUE, FALSE} Use \(O\) labelled cells as NA or as the global probability/mean estimate {TRUE, FALSE}