Skip to contents

Iterative Lasso PRS (iLasso) utilities

lasso_prs.Rd

Train, apply and visualize iterative Lasso-based PRS models with success gating.

Usage

lasso_prs(
  a1_matrix,
  divide_ratio = 0.7,
  iterative = 100,
  nfolds = 10,
  score_type = "link",
  seed = 725
)

lasso_prs_target(a1_matrix, model, lambda, score_type = "link")

lasso_prs_rank(model, rank, auc_history)

Arguments

a1_matrix

data.frame; PLINK A1 matrix.

divide_ratio

numeric; train fraction per iteration (default 0.7).

iterative

integer; number of iterations (default 100) - If set to 1, then regular lasso is performed.

nfolds

integer; CV folds for glmnet (default 10).

score_type

character; "link" (linear score, default) or "response" (probability).

seed

integer; base random seed (default 725).

model

Trained glmnet model (for target/rank functions).

lambda

Best lambda from CV.

rank

Feature frequency table.

auc_history

Tibble of iteration results.

Value

Each function returns a list:

  • lasso_prs: model, lambda, perf_train, perf_test, success_n, attempts, auc_history, feature_rank

  • lasso_prs_target: pred_df, perf, target_x, target_y

  • lasso_prs_rank: plots for feature frequency and diagnostics

See also

cv.glmnet, roc