Replicating FERRET Paper Results

Installation and Setup

You will need to install and set up FERRET and the GRN inference methods evaluated in the paper. These instructions assume a Unix/Mac environment with sudo access.

FERRET Installation and Setup

To install and set up FERRET, follow the instructions in the GitHub repository.

Python Installation and Setup

You will need Python to run scGeneRai. To install Python, run the following commands:

sudo apt-get update
sudo apt-get install python3

You will also need to install conda. To do this, run the following commands:

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
./Miniconda3-latest-Linux-x86_64.sh
export PATH=$PATH:”/home/ubuntu/miniconda3/bin”

MATLAB Installation and Setup

You will need MATLAB to run SINGE and GRISLI. To install MATLAB, do the following steps:

1. Download MATLAB. You may have institutional access. Institutional access for Harvard is available here.
2. Download the license, then select Install and Activate.
3. Run sudo unzip -X -K matlab_R2023b_glnxa64.zip -d matlab_2023b_installer, substituting version number as needed.

For GRISLI, you will also need the Bioinformatics Toolbox and the Statistics and Machine Learning Toolbox. To install these, run:

wget https://www.mathworks.com/mpm/glnxa64/mpm
chmod 777 mpm
sudo ./mpm install --release R2024a --products Bioinformatics_Toolbox Statistics_and_Machine_Learning_Toolbox

Julia Installation and Setup

You will need Julia to run PIDC. To install Julia, run the following commands:

wget https://julialang-s3.julialang.org/bin/linux/x64/1.10/julia-1.10.0-linux-x86_64.tar.gz
tar zxvf julia-1.10.0-linux-x86_64.tar.gz
vi ~/.bashrc
export PATH="$PATH:/home/ubuntu/julia-1.10.0"
sudo snap install julia --classic

R Installation and Setup

You will need R to run scSGL and LEAP. To install R, run the following commands:

sudo apt update
sudo apt install r-base r-base-dev -y

Docker Installation and Setup

You will need to install Docker to run SINGE. To install Docker, run the following commands:

sudo apt install docker.io
sudo usermod -aG docker ubuntu
docker login -u <username>

GRISLI Installation and Setup

To install GRISLI, run the following commands:

sudo apt-get install libgtk2.0-0
git clone https://github.com/PCAubin/GRISLI
wget https://files.inria.fr/spams/spams-matlab-precompiled-v2.6-2017-02-27-Linux.tar.gz
tar -zxvf spams-matlab-precompiled-v2.6-2017-02-27-Linux.tar.gz
cp -r spams-matlab-v2.6/ GRISLI/

LEAP Installation and Setup

LEAP is available from the CRAN archive.

PIDC Installation and Setup

To install PIDC, run julia on the command line and issue the following lines of code interactively:

> import Pkg
> Pkg.add(“Pkg”)
> Pkg.add("InformationMeasures")
> Pkg.add(“PyPlot”)
> Pkg.add(“LightGraphs”)
> Pkg.add(“GraphPlot”)
> Pkg.add(“NetworkInference”)

ScGeneRai Installation and Setup

1. To install scGeneRai, run the following commands:

   git clone https://github.com/PhGK/scGeneRAI.git
   sudo apt install python3-pip
   pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu
   pip install tqdm 
   cd ../scGeneRAI/
   pip install -r requirements.txt
   pip install pandas
   

2. Add the path to scGeneRai to the PYTHONPATH in .bashrc and to the PATH in .profile.

scSGL Installation and Setup

1. Run git clone https://github.com/Single-Cell-Graph-Learning/scSGL.git.
2. Run cd scSGL.
3. Check the file requirements in requirements.txt and install each one manually using pip.
4. Launch R in the command line and run install.packages("pcaPP").

SINGE Installation and Setup

To install and run SINGE using the Docker container, run the following commands:

git clone https://github.com/gitter-lab/SINGE.git
cd SINGE
wget https://hastie.su.domains/glmnet_matlab/glmnet_matlab_new.zip
unzip glmnet_matlab_new.zip -d SINGE/
sudo apt-get update
docker pull agitter/singe:0.5.1

SCORPION Installation and Setup

SCORPION is available from CRAN and GitHub

SCENIC Installation and Setup

SCENIC can be installed from GitHub using the instructions here.

Preprocessing

In most cases, you should not need to preprocess the data again. Instead, simply download them using the FERRET R package function DownloadFERRETData(). However, if you wish to replicate the preprocessing steps, the instructions for doing this are below.

Preprocessing CPTAC Data

1. Download the CPTAC data from Genomic Data Commons using the following steps:
   1. Set the Experimental Strategy to "scRNA-seq".
   2. Select "CPTAC-3" from the list of projects.
   3. Select Biospecimen > TSV and save to your machine.
   4. Select Clinical > TSV and save to your machine.
   5. Select Manifest and save to your machine.
   6. Download the scRNA-seq data using the GDC Data Transfer Tool.
2. Run CPTAC_singlecell_preprocessing.R, setting the following variables:
   • pythonpath: The path to your Python installation. This will be needed by the Reticulate R package.
   • dataDir: The path to the location where you downloaded the CPTAC data.
   • biospecimenDir: The name of the local folder (inside dataDir) where you installed the biospecimens
   • clinicalDir: The name of the local folder (inside dataDir) where you installed the clinical information
3. Run assign_pseudotime_single_cell_cptac.R to assign pseudotime based on gene expression, setting the following variables:
   • expression_dir: The path to the directory where the gene expression data are stored
   • pseudotime_result_dir: The path to the directory where you wish to store the pseudotime assignment.
4. Run assign_pseudotime_single_cell_cptac.R to assign pseudotime based on gene expression, setting the following variables:
   • expression_dir: The path to the directory where the gene expression data are stored
   • pseudotime_result_dir: The path to the directory where you wish to store the pseudotime assignment.
5. Run assign_pseudotime_single_cell_manual_cptac.R to manually assign a pseudotime starting point and endpoint based on the gene expression data and a Monocle trajectory, setting the following variables:
   • expression_dir: The path to the directory where the gene expression data are stored
   • pseudotime_result_dir: The path to the directory where you wish to store the pseudotime assignment.
6. Run single_cell_PC_expression_CPTAC_subset.R to project CPTAC gene data onto the first 100 principal components, subset the number of cells, and split into folds, setting the following variables:
   • pca_result_dir: The file path where you wish to store the results of PCA
   • plot_result_dir: The file path where you wish to store the PCA plots
   • expression_dir: The path to the file containing the preprocessed expression data
   • split_dir: The path to the directory where you wish to store the data after splitting into folds
7. Run single_cell_CPTAC_splits_subset_pseudotime_only.R to run the corresponding splits for the assigned pseudotime trajectory, setting the following variables:
   • split_result_dir: The path to the directory where you wish to store the data after splitting into folds
   • expression_dir: The path to the file containing the preprocessed expression data
   • split_result_dir_pseudotime: The path to the directory where you wish to store the pseudotime data after splitting into folds

Preprocessing HTAN Data

1. Download the HTAN SCLC data from the HTAN website and accompanying metadata using the following filters:
   • Organ = Lung NOS
   • Atlas = MSK
   • Assay = scRNA-seq
   • File Format = csv
2. Remove the following samples, which do not correspond to lung tissue:
   • RU 681 and RU 255 (brain metastasis)
   • RU 666 (bone metastasis)
   • MRM35429319 and 35379369 (spinal fluid)
   • 1559_1262C P96 and RU1141 (not listed in metadata file)
   • RU1128 (liver metastasis)
3. Run preprocess_htan_msk.R, setting the following variables:
   • dataDir: The path to the directory where the data were downloaded
4. Run assign_pseudotime_single_cell_htan_msk.R, setting the following variables:
   • dataDir: The path to the directory where the data were downloaded
5. Run assign_pseudotime_manual_single_cell_htan_msk.R, setting the following variables:
   • pseudotime_result_dir: The path to the directory where the initial pseudotime assignments were stored
6. Run single_cell_HTAN_splits_subset_all_cells.R, setting the following variables:
   • expression_dir: The path to the file containing the preprocessed expression data
   • split_result_dir: The path to the directory where you wish to store the data after splitting into folds
   • pseudotime_dir: The path to the directory with the final pseudotime assignments
   • pseudotime_result_dir: The path to the directory where you wish to store the pseudotime data after splitting into folds

ChIP-seq Data Evaluation

Download Data

1. Navigate to CISTROME DB
2. Select the following filters:
   • Species = Homo sapiens
   • Biological Sources = B Lymphocyte, B-lympocytes, fibrobasts (sic), Fibroblast
3. Download only those files meeting the following criteria:
   • Passing all quality controls
   • Genome is known and is not hg18
   • Autosomes are available
   • The factor is available for both fibroblasts and B-lymphocytes
   • The cells are untreated
4. Download the GENCODE genomes for hg19 and hg38.

Preprocess Data

1. Run make_TSS_BED.R to create a BED file with TSS from the reference genomes, setting the following variables:
   • hg19Loc: The path to the hg19 reference genome
   • hg38Loc: The path to the hg38 reference genome
   • hg19BedLoc: The path to the file where you wish to store the hg19 BED file
   • hg38BedLoc: The path to the file where you wish to store the hg38 BED file
2. Download bigWigToWig.
3. Run the following for each file in BigWig format to convert it to Wig format:

./bigWigToWig source.bw dest.wig

4. Download wig2bed
5. Run the following for each file in Wig format to convert it to BED format:

wig2bed < source.wig | bedmap --echo --max-element -> dest.bed

6. Run fixFormat.R to fix the format of two BED files, setting the following variables:
   • GSM1869138_BJ_PolII_MeDiChi-seq_peaks: The path to the file containing the sample with this name
   • GSM1869138_BJ_PolII_MeDiChi-seq_peaks_out: The path to the file where you wish to store the version of this sample with the fixed format
   • GSM1869150_BJELM_PolII_MeDiChi-seq_peaks: The path to the file containing the sample with this name
   • GSM1869150_BJELM_PolII_MeDiChi-seq_peaks_out: The path to the file where you wish to store the version of this sample with the fixed format
7. Install bedtools. To do this on a Mac, run the following commands:

brew tap homebrew/science
brew install bedtools

9. Run the following to find the promoter binding sites for each file:

bedtools intersect -wo -a promoters.bed -b peaks.bed > promoter_binding_sites.bed

Generate ChIP-seq Networks

Run build_ChIPseq_networks.R, setting the following variables:

• baseRepo: The path to the parent directory where the fibroblast and B lymphocyte BED files are stored

Evaluate Networks Using FERRET

Run evaluate_ChIPseq.R, setting the following variables:

• networkDir: Path to directory where the networks are stored
• resultDir: Path to directory where you wish to store the FERRET results
• pathwayFile: Path to the pathway file downloaded from MSigDB
• differentialPathwaysDir: Path to the directory where you wish to save the differential pathways

Random Network Evaluation

1. Run run_FERRET_on_random_networks.R, setting the following variables:
   • inputDir: The path to the directory where you wish to store the random networks
   • outputDir: The path to the directory where you wish to store the FERRET results
2. Run consolidate_AUC_random.R, setting the following variables:
   • outputDir: The path to the directory where the FERRET results from the original random networks are stored
3. Run run_FERRET_on_random_networks_bipartite.R, setting the following variables:
   • inputDir: The path to the directory where the random networks from step 1 are stored
   • outputDir: The path to the directory where you wish to store the FERRET results
4. Run consolidate_AUC_random.R, setting the following variables:
   • outputDir: The path to the directory where the FERRET results from the bipartite random networks are stored

CPTAC FERRET Evaluation

Running GRN Inference Methods

GRISLI

1. Run transpose_input.R to transpose the data, setting the following variables:
   • split_dir: The path to the directory containing the splits
   • transpose_split_dir: The path to where you wish to store the transposed split files
2. In the file run_grisli_single.m, change the following line as needed to include the installation path:

   addpath('/home/ubuntu/GRISLI/GRISLI/');
   

3. Run ./run_grisli_parallel.sh, which calls run_grisli_single.m, setting the following variables:
   • EXPRESSION_DIR: The path to the directory containing the transposed expression data
   • PSEUDOTIME_DIR: The path to the directory containing the pseudotime data
   • OUTPUT_DIR: The path to the directory where you want the output GRNs to be saved
   • LOGGING_DIR: The path to the directory where you want the logs to be saved
   • GRISLI_LOCAL: The path to the directory where GRISLI is installed

PIDC

Run ./run_PIDC_parallel.sh, which calls runPIDCsingle.jl, setting the following variables:

• EXPRESSION_DIR: The path to the directory containing the transposed expression data
• OUTPUT_DIR: The path to the directory where you want the output GRNs to be saved
• LOGGING_DIR: The path to the directory where you want the logs to be saved

scSGL

Run ./run_scsgl_parallel.sh, setting the following variables:

• EXPRESSION_DIR: The path to the directory containing the expression data
• OUTPUT_DIR: The path to the directory where you want the output GRNs to be saved
• LOGGING_DIR: The path to the directory where you want the logs to be saved

scGeneRai

Run ./run_scgenerai_parallel.sh, which calls scgenerai_run_single.py, setting the following variables:

• EXPRESSION_DIR: The path to the directory containing the expression data
• OUTPUT_DIR: The path to the directory where you want the output GRNs to be saved
• LOGGING_DIR: The path to the directory where you want the logs to be saved

LEAP

Run ./run_LEAP_parallel.sh, which calls run_LEAP_single.py, setting the following variables:

• EXPRESSION_DIR: The path to the directory containing the transposed expression data
• PSEUDOTIME_DIR: The path to the directory containing the pseudotime data
• OUTPUT_DIR: The path to the directory where you want the output GRNs to be saved
• LOGGING_DIR: The path to the directory where you want the logs to be saved

SINGE

1. Run make_mat_files_singe.R, setting the following variables:
   • expressionPath: The path to the transposed expression data
   • pseudotimePath: The path to the pseudotime data
   • outputPath: The path where you wish to save the formatted data
2. Run ./run_singe_parallel.sh, setting the following variables:
   • EXPRESSION_DIR: The path to the directory containing the expression data formatted for SINGE
   • SINGE_DIR: The path to the directory where SINGE is installed
3. Run convert_mat_to_csv.R, setting the following variables:
   • matPath: The path where SINGE is installed
   • outputPath: The path where you wish to save the CSV files

Evaluating Performance Using FERRET

1. Run format_PIDC_results_for_FERRET.R, setting the following variables:
   • originalDir: The directory containing the original PIDC results
   • consolidatedDir: The directory containing the reformatted PIDC results
2. Run run_FERRET.R, setting the following variables:
   • scGeneRai_results: The path to the directory containing networks inferred using scGeneRai
   • scGeneRai_results_formatted: The path to the directory where you wish to store the reformatted scGeneRai results
   • GRISLI_results: The path to the directory containing the networks inferred using GRISLI
   • scSGL_results: The path to the directory containing the networks inferred using scSGL
   • PIDC_results: The path to the directory containing the networks inferred using PIDC
   • LEAP_results: The path to the directory containing the networks inferred using LEAP
   • SINGE_results: The path to the directory containing the networks inferred using SINGE
   • roc_auc_parent_dir: The path to the parent directory where the FERRET results for each inference method will be stored

Evaluating Pathways

1. Run run_FERRET_pathways.R, setting the following variables:
   • scGeneRai_results: The path to the directory containing networks inferred using scGeneRai
   • scGeneRai_results_formatted: The path to the directory where you wish to store the reformatted scGeneRai results
   • GRISLI_results: The path to the directory containing the networks inferred using GRISLI
   • scSGL_results: The path to the directory containing the networks inferred using scSGL
   • PIDC_results: The path to the directory containing the networks inferred using PIDC
   • LEAP_results: The path to the directory containing the networks inferred using LEAP
   • SINGE_results: The path to the directory containing the networks inferred using SINGE
   • pathway_dir: The path to the parent directory where the pathway analysis results should be stored
   • GMT: The path to the GMT pathway file
2. Run pathway_frequency_CPTAC_cellTypePairs.R, to compute the frequency of pathway enrichment when compared across two cell types, setting the following variables:
   • pathway_dir_GRISLI: The path to the directory containing the pathway results for GRISLI
   • pathway_dir_LEAP: The path to the directory containing the pathway results for LEAP
   • pathway_dir_PIDC: The path to the directory containing the pathway results for PIDC
   • pathway_dir_scGeneRai: The path to the directory containing the pathway results for scGeneRai
   • pathway_dir_scSGL: The path to the directory containing the pathway results for scSGL
   • pathway_dir_SINGE: The path to the directory containing the pathway results for SINGE
   • pathway_dir_GRISLI_all: The path to the directory containing the consolidated pathway results for GRISLI
   • pathway_dir_LEAP_all: The path to the directory containing the consolidated pathway results for LEAP
   • pathway_dir_PIDC_all: The path to the directory containing the consolidated pathway results for PIDC
   • pathway_dir_scGeneRai_all: The path to the directory containing the consolidated pathway results for scGeneRai
   • pathway_dir_scSGL_all: The path to the directory containing the consolidated pathway results for scSGL
   • pathway_dir_SINGE_all: The path to the directory containing the consolidated pathway results for SINGE
   • pathway_dir_GRISLI_table: The path to the directory containing the consolidated pathway table for GRISLI
   • pathway_dir_LEAP_table: The path to the directory containing the consolidated pathway table for LEAP
   • pathway_dir_PIDC_table: The path to the directory containing the consolidated pathway table for PIDC
   • pathway_dir_scGeneRai_table: The path to the directory containing the consolidated pathway table for scGeneRai
   • pathway_dir_scSGL_table: The path to the directory containing the consolidated pathway table for scSGL
   • pathway_dir_SINGE_table: The path to the directory containing the consolidated pathway table for SINGE

Making Bar Plots

Run consolidate_AUC_by_cell_type_CPTAC.R, changing the following variables:

• rocAucDirScSGL: Path to the directory storing FERRET results for scSGL
• rocAucDirPIDC: Path to the directory storing FERRET results for PIDC
• rocAucDirLEAP: Path to the directory storing FERRET results for LEAP
• rocAucDirScGeneRai: Path to the directory storing FERRET results for scGeneRai
• rocAucDirSINGE: Path to the directory storing FERRET results for SINGE
• rocAucDirGRISLI: Path to the directory storing FERRET results for GRISLI
• ranges_all_file: Path to the file where you wish to store AUC ranges
• jaccard_plot: Path to the PDF file where you wish to plot Jaccard bars
• indegree_plot: Path to the PDF file where you wish to plot In-Degree bars
• outdegree_plot: Path to the PDF file where you wish to plot Out-Degree bars

HTAN FERRET Evaluation

Running GRN Inference Methods

SCORPION

1. Download Grand TF binding and PPI files.
2. Modify the following run_scorpion.R variables:
   • motif_file: Path to the TF binding motif file
   • ppi_file: Path to the PPI file
3. Run ./run_scorpion_HTAN.sh, setting the following variables:
   • EXPRESSION_DIR: The path to the directory where the expression data are stored
   • OUTPUT_DIR: The path to the directory where you wish to store the output
   • LOGGING_DIR: The path to the directory where you wish to store the logs

SCENIC

1. In R, launch the following commands to install SCENIC and save the SCENIC RMD file:

if(!require("SCENIC")){
  devtools::install_github("aertslab/SCENIC") 
}
vignetteFile <- file.path(system.file('doc', package='SCENIC'), "SCENIC_Running.Rmd")
file.copy(vignetteFile, "SCENIC_myRun.Rmd")

2. Download the motif database.
3. Modify the following variables in run_scenic.R:
   • cisTarget_database_dir: Path to directory where you downloaded the motif database.
4. Run ./run_scenic_HTAN.sh, setting the following variables:
   • EXPRESSION_DIR: The path to the directory where the expression data are stored
   • OUTPUT_DIR: The path to the directory where you wish to store the output

Evaluating Performance Using FERRET

Run run_FERRET_HTAN_AWS.R, setting the following variables:

• networks_SCORPION_file: The path to the directory containing the SCORPION networks
• networks_SCENIC_file: The path to the directory containing the SCENIC networks

Evaluating Pathways

1. Run make_pathways_plot_HTAN.R, setting the following variables:
   • GMT: The path to the GMT pathway file
   • network_dir_SCORPION: The path to the directory that has the SCORPION networks
   • network_dir_SCENIC: The path to the directory that has the SCENIC networks
   • pathway_dir_SCORPION: The path to the directory where you wish to store the SCORPION pathways
   • pathway_dir_SCENIC: The path to the directory where you wish to store the SCENIC pathways
2. Run pathway_frequency_HTAN_cellTypePairs.R, to compute the frequency of pathway enrichment when compared across two cell types, setting the following variables:
   • pathway_dir_SCORPION: The path to the directory containing the pathway results for SCORPION
   • pathway_dir_SCENIC: The path to the directory containing the pathway results for SCENIC
   • pathway_dir_SCORPION_all: The path to the directory containing the consolidated pathway results for SCORPION
   • pathway_dir_SCENIC_all: The path to the directory containing the consolidated pathway results for SCENIC
   • pathway_dir_SCORPION_table: The path to the directory containing the consolidated pathway table for SCORPION
   • pathway_dir_SCENIC_table: The path to the directory containing the consolidated pathway table for SCENIC

Making Bar Plots

Run consolidate_AUC_by_cell_type_HTAN.R, changing the following variables:

• rocAucDirSCORPION: Path to the directory storing FERRET results for SCORPION
• rocAucDirSCENIC: Path to the directory storing FERRET results for SCENIC
• ranges_all_file: Path to the file where you wish to store AUC ranges
• jaccard_plot: Path to the PDF file where you wish to plot Jaccard bars
• indegree_plot: Path to the PDF file where you wish to plot In-Degree bars
• outdegree_plot: Path to the PDF file where you wish to plot Out-Degree bars

Running Time Evaluation for HTAN

GRISLI

Run ./run_grisli_testsingle_HTAN.sh, setting the following variables:

• EXPRESSION_DIR: The path to the directory where the expression data are stored
• PSEUDOTIME_DIR: The path to the directory where the pseudotime data are stored
• OUTPUT_DIR: The path where you wish to store the output
• GRISLI_DIR: The path where GRISLI is stored

LEAP

Run ./run_LEAP_testsingle_HTAN.sh, setting the following variables:

• EXPRESSION_DIR: The path to the directory where the expression data are stored
• PSEUDOTIME_DIR: The path to the directory where the pseudotime data are stored
• OUTPUT_DIR: The path where you wish to store the output

PIDC

Run ./run_PIDC_testsingle_HTAN.sh, setting the following variables:

• EXPRESSION_DIR: The path to the directory where the expression data are stored
• OUTPUT_DIR: The path where you wish to store the output

SCENIC

Run ./run_SCENIC_testsingle_HTAN.sh, setting the following variables:

• EXPRESSION_DIR: The path to the directory where the expression data are stored
• OUTPUT_DIR: The path where you wish to store the output

scGeneRai

Run ./run_scgenerai_testsingle_HTAN.sh, setting the following variables:

• EXPRESSION_DIR: The path to the directory where the expression data are stored
• OUTPUT_DIR: The path where you wish to store the output

SCORPION

Run ./run_scorpion_testsingle_HTAN.sh, setting the following variables:

• EXPRESSION_DIR: The path to the directory where the expression data are stored
• OUTPUT_DIR: The path where you wish to store the output

scSGL

Run ./run_scsgl_testsingle_HTAN.sh, setting the following variables:

• EXPRESSION_DIR: The path to the directory where the expression data are stored
• OUTPUT_DIR: The path where you wish to store the output