Benchmark of Differential TF activity methods

This repository contains the code used in Gerbaldo, Sonder et al., for the benchmark of differential TF activity methods based on ATAC-seq data.

Repository structure

• fullFrags: benchmark on real datasets, using all fragments.
• nucFree: benchmark on real datasets, using only nucleosome-free fragments.
• fullOnNucFreePeaks: benchmark on real datasets, using counts of all fragments on nucleosome-free peaks.
• withArchetypes: benchmark on real datasets, using motif archetypes.
• Scripts: wrappers and higher-level functions.
• Figures: code behind the paper's figures.
• simulations: code for the generation of the semi-simulated datasets, and the benchmarks performed on those.
• singleCell: code for the cell-level analysis.
• misc: misc scripts preparing data objects for other analyses/plotting.
• TRAFTAC: code for the analysis of the TRAFTAC dataset.

Main benchmark folders

Each of the 3 main benchmark folders contains one folder per dataset, with a peaks and a seq_files subfolders containing, respectively, the merged peaks and the (aligned) reads/fragments. Due to size limitations the reads are not included in the repository, but are available on request.

To re-run the benchmark, simply do, from one of these folders:

source("../Scripts/runMethods.R")
datasets <- getDatasets()
runAll(datasets)

To run only the top methods, use:

runAll(datasets, methods=getMethods(onlyTop=TRUE))

Results can be benchmarked using:

source("../Scripts/compileBenchmark.R")
interactors <- readRDS("../Scripts/allInteractors.rds")
res <- compileBenchmark(datasets, interactors=interactors)

The precompiled results are also available in the respective folders.