Skip to contents

Run Association Analysis

Source: R/run_association.R
run_association.Rd

Perform GLM-based association testing between a specified outcome and features from exposures, omics, latent factors, or GO PCs. Automatically adjusts for covariates and supports both Gaussian and binomial models.

Usage

run_association(
  expomicset,
  outcome,
  source = c("omics", "exposures", "factors"),
  covariates = NULL,
  feature_set = NULL,
  log_trans = TRUE,
  top_n = NULL,
  family = "gaussian",
  correction_method = "fdr",
  action = "add",
  min_genes = 10,
  feature_col = NULL,
  mirna_assays = NULL
)

Arguments

expomicset

A MultiAssayExperiment object containing data and metadata.

outcome

The outcome variable name (must be in colData).

source

Source of features to test. One of "omics", "exposures", "factors".

covariates

Optional vector of covariate names to include in the model.

feature_set

Optional character vector of exposure or GO terms to test.

log_trans

Optional boolean value dictating whether or not to log2 transform omics features.

top_n

Optional integer: if using omics source, select top n most variable features.

family

GLM family; "gaussian" or "binomial".

correction_method

Method for p-value adjustment (default: "fdr").

action

If "add" (default), saves results to metadata; else returns results as list.

min_genes

Minimum number of genes required to compute GO PCs.

feature_col

If using GO PCs, the column in rowData for matching gene symbols or IDs.

mirna_assays

Optional character vector of assays to exclude when extracting GO terms.

Value

If action = "add", returns updated MultiAssayExperiment. Otherwise, returns a list of:

  • results_df: tidy summary of associations

  • covariates: the covariates used

  • model_data: model matrix used in the GLMs

Examples


#' # create example data
mae <- make_example_data(
    n_samples = 10,
    return_mae = TRUE
)
#> Ensuring all omics datasets are matrices with column names.
#> Creating SummarizedExperiment objects.
#> Creating MultiAssayExperiment object.
#> MultiAssayExperiment created successfully.

# run association tests
mae <- mae |>
    run_association(
        source = "exposures",
        feature_set = c("exposure_pm25", "exposure_no2"),
        outcome = "smoker",
        covariates = c("age"),
        family = "binomial"
    )
#> Running GLMs.
#> Warning: glm.fit: algorithm did not converge
#> Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred