CombiROC is a totally new music in multi-markers analysis: an R package for efficient and easy combinatorial selection of biomarkers and sensitivity/specificity-driven prioritization of features.
This is the development version of CombiROC package (combiroc), code in this repo is work in progress and it is uploaded here “as-is” with no warranties implied. Improvements and new features will be added on a regular basis, please check on this github page for new features and releases.
The CombiROC approach was first published by Mazzara et al. Scientific Reports 2017 as a Shiny app. A description of the analitycal protocol is also published in Bombaci & Rossi, Methods Mol Biol 2019. The Shiny web-app version of CombiROC is still available at combiroc.eu, but it has limited features (as well as low computational power) and is not further maintained. For full capabilities, new and improved features and customized analyses is advisable to install this package on your own machines.
# First, install remotes from CRAN and run it
install.packages("remotes")
library(remotes)
# Then install the development version of CombiROC from GitHub
remotes::install_github("ingmbioinfo/combiroc",
dependencies = TRUE, build_vignettes = TRUE)
# remotes is a lightweight replacement of install functions from devtools
# if you already have devtools, you can also use devtools::install_github() Full documentation is in the package’s vignette. You can also find the rendered version of the vignette in the combiroc-package website created with pkgdown.
library(combiroc)
# load the preformatted demo dataset
# (you can load a dataset of yours using load_data() function: see full docs)
data <- demo_data
# shape it in long format (prone to plotting)
data_long <- combiroc_long(data)
# study the distribution of you markers' signal
# arguments values to be adjusted according to data
distr <- markers_distribution(data_long, case_class = 'A',
y_lim = 0.0015, x_lim = 3000,
signalthr_prediction = TRUE,
min_SE = 40, min_SP = 80,
boxplot_lim = 2000)
# explore the distr object: boxplot of signals
distr$Boxplot
# explore the distr object: densities of classes with signal threshold (signalthr)
distr$Density_plot
distr$Density_summary
# explore the distr object: ROC and its coordinates
distr$ROC
head(distr$Coord, n=10)
# combinatorial analysis
tab <- combi(data, signalthr = 407, combithr = 1)
# SE & SP computation
mks <- se_sp(data, tab)
# ranked combinations
rmks <- ranked_combs(data, mks, case_class = 'A', min_SE = 40, min_SP = 80)
# check ranked combinations
rmks$table
rmks$bubble_chart
# results report for specific markers/combinations
reports <-roc_reports(data, markers_table = tab, case_class = 'A',
single_markers =c('Marker1'),
selected_combinations = c(11,15))
# results outputs
reports$Plot
reports$MetricsIf you find a bug, or to share ideas for improvement, feel free to start an issue. We do have a roadmap but we also listen!
We were so happy to finally had the chance to develop the combiroc package that we felt very “rock”: this is why the combiroc hexagon sticker logo is a homage to Eddie Van Halen who left us in 2020, and the “Frankenstrat”, his iconic guitar.