Skip to content

Repository for: PEDF, a Pleiotropic WTC-LI Biomarker: Machine Learning Biomarker Identification and Validation

Notifications You must be signed in to change notification settings

NolanLabNYU/PEDF-a-pleiotropic

Repository files navigation

PEDF-a-pleiotropic

catalogue of files in repo

Custom Functions

==> clean.summary.R <==
clean.summary <- function(arg1, warn=TRUE){ ##takes output of summary() and returns an html table with median(iqr) for continuous (numeric) variables

==> fix.peak.labels.R <==
fix.peak.labels <- function(arg1) { #takes a vector of strings of metabolite/peak names (that were altered by R's makenames()) #and fixes names to match the format in data sheets

==> ham.dist.R <==
ham.dist <- function(arg1) { ##function ham.dist computes all pairwise hamming distances among the columns of arg1, a matrix

==> parseSlurmNodeList.R <==
parseSlurmNodeList<- function(arg1) { #parses SLURM_NODELIST environment variable into a list of nodes (i.e. to be passed to makePSOCKcluster() or h2o.init())

==> uniqueEl.R <==
uniqueEl <- function(arg1) { ##computes the number of unique elements in a pair of sets, for all pairs of columns in a matrix

==> permimp <==
customized version of permimp to support parallel processing on SLURM modifications were made to /R/doPermimp.R, DESCRIPTION, NAMESPACE, and MD5 to build, use: R CMD build permimp --no-build-vignettes

Analysis Scripts

==> compile_met_info_and_results.R <==
compile variable information and results, save for graphing, calculate correlation for variable importance rankings

==> cpi.R <==
cpi.R measure cpi using custom parallel permimp on hpc cluster

==> plot_tune_results.R <==
graph tune results

==> regressions.R <==
regressions for validation cohort

==> rf_tune.R <==
rf hyperparameter search/tune

==> clustergram_code.m <==
code to reproduce clustergram

Data Files

==> roc_coords.csv <==
contains coordinates for ROC curves

==> met_info.sav <==
contains classical permutation importance and conditional permutation importance as described in paper can be read into R using haven::read_sav("met_info.sav")

==> tune.results.csv <==
contains results of tuning process for random forests with all variables

==> tune.results.refined.csv <==
contains results of tuning process for random forests with refined variable profile

==> lasso.results.csv <==
contains results of lasso tuning procedure

About

Repository for: PEDF, a Pleiotropic WTC-LI Biomarker: Machine Learning Biomarker Identification and Validation

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published