Running vertex-wise linear mixed models

library(verywise)
#> Welcome, verywise user!
#> This is version: 0.1.3
#> For questions, issues, and bug reports, please see https://github.com/SereDef/verywise

The main function in the verywise package is run_vw_lmm which fits a vertex-wise linear mixed model to the data. Here is an example:

# Run a linear mixed model
run_vw_lmm(
  formula = vw_thickness ~ sex * age + site + (1 | id), # model formula
  pheno = phenotype, # An R object already in memory or a path to file
  subj_dir = "./VeryWiseUser/SimulatedExample", # Neuro-imaging data location
  hemi = "both", # (default) which hemispheres to run
  n_cores = 4 # number of cores for parallel processing
  ...
)

The function will first check the inputs and then run lme4::lmer on every vertex in both hemispheres.

The most important parameters you need to know about:

formula: the model formula specifying the linear mixed model. This uses lme4 syntax.
pheno : either the phenotype data object (already loaded in the global environment) or a string containing a file path. Supported file extensions are: rds, csv, txt and sav.
subj_dir : a string containing the path to the FreeSurfer data, this expects a verywise structure.

You can optionally also specify:

outp_dir: a string containing the path, where do you want results to be stored. If none is provided, a “results” sub-directory will be created inside subj_dir.
hemi: which hemispheres to run (default = “both”)
seed: (default = 3108) random seed.
n_cores: (default = 1) number of cores for parallel processing.
FS_HOME: FreeSurfer directory, i.e. $FREESURFER_HOME.

Check out the output directory.

Serena Defina

2025-03-22