Skip to contents

Compute isoTWAS model for a set of isoforms/transcripts

compute_isotwas.Rd

The function runs multivariate regression models for a set of transcripts and outputs the best model based on cross-validated R-squared.

Usage

compute_isotwas(
  X,
  Y,
  gene_exp = NULL,
  Y.rep,
  R = NULL,
  id,
  omega_est = "replicates",
  omega_nlambda = 10,
  method = c("mrce_lasso", "multi_enet", "joinet", "spls", "sgl", "mtlasso", "stacking",
    "graph_reg", "univariate"),
  similarity_matrix = NULL,
  predict_nlambda = 20,
  family = "gaussian",
  scale = FALSE,
  alpha = 0.5,
  nfolds = 5,
  verbose = TRUE,
  par = FALSE,
  n.cores = NULL,
  tx_names = NULL,
  seed = NULL,
  run_all = TRUE,
  return_all = FALSE,
  tol.in = 0.001,
  maxit.in = 1000
)

Arguments

X

matrix, design matrix of SNP dosages

Y

matrix, matrix of G isoform expression across columns

gene_exp

vector, vector of total gene expression

Y.rep

matrix, matrix of G isoform expression with replicates

R

int, number of replicates

id

vector, vector of sample ids showing rep to id

omega_est

character, 'replicates' or 'mean' to use Y.rep or Y

omega_nlambda

int, number of omegas to generate

method

character, vector of methods to use. Available methods:

  • "mrce_lasso": MRCE with lasso penalty

  • "multi_enet": Multivariate elastic net

  • "joinet": Joinet stacked elastic net

  • "spls": Sparse partial least squares

  • "sgl": Sparse group lasso

  • "mtlasso": Multi-task lasso (L21 regularization)

  • "stacking": Super learner stacking of multiple methods

  • "graph_reg": Graph-regularized regression (requires similarity_matrix)

  • "univariate": Best of univariate elastic net, BLUP, SuSiE

similarity_matrix

matrix, optional similarity matrix for graph_reg method. Can be created using similarity_from_gtf or create_similarity_from_exons. If NULL, graph_reg is skipped.

predict_nlambda

int, number of lambdas in MRCE

family

character, glmnet family

scale

logical, T/F to scale Y by Omega

alpha

numeric, elastic net mixing parameter

nfolds

int, number of CV folds

verbose

logical, print progress messages

par

logical, uses mclapply to parallelize model fit

n.cores

int, number of parallel cores

tx_names

vector, character vector of tx names - order of columns of Y

seed

int, random seed

run_all

logical, run all methods

return_all

logical, return R2 for all models?

tol.in

numeric, tolerance for objective difference

maxit.in

int, maximum number of iterations

Value

optimal isoTWAS model