Multivariate Sparse Group LASSO — multivariate

Fits a sparse group LASSO model for multivariate isoform prediction. This method encourages both: - Group-level sparsity: same SNPs tend to affect multiple isoforms (shared effects) - Within-group sparsity: isoform-specific effects

Usage

multivariate_sgl(
  X,
  Y,
  alpha = 0.5,
  nlambda = 20,
  lambda_min_ratio = 0.01,
  nfolds = 5,
  standardize = FALSE,
  verbose = FALSE,
  seed = 123,
  par = FALSE,
  n.cores = 1
)

Arguments

X: matrix, design matrix of SNP dosages
Y: matrix, matrix of G isoform expression across columns
alpha: numeric, mixing parameter between group and individual sparsity (0-1) alpha=1 is pure LASSO, alpha=0 is pure group LASSO
nlambda: int, number of lambda values to try
lambda_min_ratio: numeric, ratio of lambda_min to lambda_max
nfolds: int, number of CV folds
standardize: logical, standardize X before fitting
verbose: logical
seed: int, random seed
par: logical, use parallel processing for CV folds
n.cores: int, number of cores for parallel processing

Value

isotwas_model object

Details

The penalty is: alpha * ||B||_1 + (1-alpha) * sum_j ||B_j||_2 where B_j is the j-th row of B (effects of SNP j across all isoforms)