Skip to content

Latest commit

 

History

History
141 lines (113 loc) · 10.1 KB

README.md

File metadata and controls

141 lines (113 loc) · 10.1 KB
Raw

scTIGER (Single-cell Temporal Inference of Gene Regulatory Networks)

The Single-cell Temporal Inference of Gene Regulatory (scTIGER) Networks method is a computational method designed to predict gene regulatory networks (GRNs) using paired datasets of case versus control experiments. After constructing a gene co-differential expression network, scTIGER employs cell-based pseudotiming, an attention-based convolutional neural network method, and permutation-based significance testing for inferring GRNs among gene modules.

Corresponding Paper: Dautle M, Zhang S, Chen Y. scTIGER: A Deep-Learning Method for Inferring Gene Regulatory Networks from Case versus Control scRNA-seq Datasets. International Journal of Molecular Sciences. 2023; 24(17):13339. https://doi.org/10.3390/ijms241713339 MyImage

Functionality

We provide a YouTube Video which provides details about how to apply the pipeline to your own data and the potential uses for our pipeline.

  • Cell- and condition-specific GRN inference from paired scRNA-seq datasets
    • High precision and low false positive rates
    • Highly robust against dropout noise

Prerequisites

General

  • Linux
  • Python
  • PyTorch
  • Optional: CUDA

Required Python Packages

  • numpy
  • pandas
  • matplotlib
  • networkx
  • scipy
  • scanpy
  • leidenalg
  • bambi
  • arviz
  • numba
  • All packages with versions from our successful running of scTIGER can be found here

Installation

We provide a YouTube Video to help you install scTIGER. The steps vary slightly as package updates have been made since we created the installation video, but the steps below can be used in the same fashion as the video details to install and use scTIGER.

We recommend installing the packages using a conda environment. Information on downloading Anaconda can be found here. You may use the following steps to install the necessary packages to a new environment. Note there is a different installation process for CUDA capable machines.

General installation:

conda create -n scTIGER python=3.11
conda activate scTIGER
git clone https://github.com/chenyongrowan/scTIGER
cd scTIGER
chmod +x run_scTIGER.py
unzip Data/ProstateCancer/Patient4_Benign_endothelial.zip
conda install pytorch==2.1.2 torchvision torchaudio cpuonly==2.0 -c pytorch
conda install -c madison.dautle sctiger

CUDA capable installation:

conda create -n scTIGER python=3.11
conda activate scTIGER
git clone https://github.com/chenyongrowan/scTIGER
cd scTIGER
chmod +x run_scTIGER.py
unzip Data/ProstateCancer/Patient4_Benign_endothelial.zip
conda install pytorch==2.1.2 torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia
conda install -c madison.dautle sctiger

Data

Required: Two scRNA-seq datasets, one case and one control Format: A non-normalized CSV file with genes as rows and cells as columns. Genes must be in first column of CSV file

Note: Case and control datasets should either be the same cell type and two different experimental conditions OR the same experimental condition and two different cell types

Sample Datasets

There are multiple sample datasets under the Data folder.

  1. The ProstateCancer1 folder contains datasets for one patient. The files were processed to contain only one cell type. They are also separated into benign and tumor cells.
  2. The RemoteMemoryFormation folder contains preprocessed datasets. Datasets contain only neurons. Only fear conditioned (FC) and controls were selected for the Chen2 dataset.

Running

Preprocessing

We have included a file (10x_preproccess.py) to convert the resulting files of a 10x sequencing to the gene expression matrix required for scTIGER. It has a required flag (-d/--directoryPath) which takes the path to a directory with the following files which are output from 10x sequencing:

  • features.tsv.gz
  • matrix.tsv.gz
  • barcodes.tsv.gz

This script will output the gene expression matrix for those files which can be inputted into scTIGER. The command to run this script should follow the following format:

./10x_preprocess.py -d ./Path_to_dir_with_10x_files

scTIGER

The main folder of this repository contains two main files, scTIGER.py which contains the definitions and functions utilized in run_scTIGER.py which is the script to run scTIGER.

scTIGER is set up as a single-line command with the following flags:

Required flags

  • -goi/--geneOfInterest: One or more genes of interest. Separate multiple genes with a "+" (ex. Arc+Bdnf)
  • -ctrl/--control: Path to the csv file containing control cells. Provide file with cells as columns and genes as rows. The gene names should be the first column in the file. The file must contain at least 10 cells
  • -exp/--experimental: Path to the csv file containing the case cells. Provide file with cells as columns and genes as rows. The gene names should be the first column in the file. The file must contain at least 10 cells

Optional flags

  • -p/--permutations: Number of permutations to run. Default 100
  • -top/--numTopGenes: Number of top correlated genes selected. Default 50
  • -zero/--zeroThresh: Threshold for number of 0's tolerated for a gene. Default 0.30
  • -t/--timesteps: Maximum number of steps allowed between interactions. 0 implies a direct, causal interaction. 1 implies one interaction between the source and target gene. Default is 0.
  • -s/--start: Starting point for scTIGER. Default 1 (Run scTIGER and and generate GRN files). 2 uses existing scTIGER output to generate GRN visualization files if you'd like to change the alpha level.
  • --cuda: CUDA use on when flag included. Leave flag out if using CPU based discovery
  • -o/--output: Output directory name. Default is 'scTIGER_Output'
  • -a/--alpha: Alpha value for determining significant gene interactions by scTIGER discovery. Default 0.05

The general command should appear as:

./run_scTIGER.py -goi Gene1+Gene2+Gene3 -ctrl ./path_to_file -exp ./path_to_file

The optional flags can be added to change default parameters as follows:

./run_scTIGER.py -goi Gene1+Gene2+Gene3 -ctrl ./path_to_file -exp ./path_to_file -p 50 -top 90 -zero 0.20 -t 1 --cuda -o NameOfDirectory_Output -s 2

The flags can appear in any order, not just the order detailed here.

To run a sample dataset from the scTIGER folder, use the following command AFTER unzipping Patient4_Benign_endothelial.zip:

./run_scTIGER.py -goi AR+PTEN+ERG -ctrl ./Data/ProstateCancer/Patient4_Benign_endothelial.csv -exp ./Data/ProstateCancer/Patient4_Tumor_endothelial.csv -p 50 -top 100 -zero 0.15 -o SampleResult_ProstateCancer

We have also included an Example folder which allows users to run the example to make sure their download of scTIGER is functioning. To use it, simply download the scTIGER package, make the Example folder your working directory, and run ./Example/runExample.py in your terminal. It will output the SampleResult_ProstateCancer directory. The example does not run scTIGER with CUDA. (Be sure to unzip the data in the Example folder first).

This directory contains sample output for scTIGER. The main folder contains the raw counts for each gene of interest entered, a .txt file containing the details of the command for later reference, and 3 directories. The Graphs directory contains histograms displaying the number of interactions detected with a particular percentage recovery as a function of the percent recovery for each gene. The GRN_Visualization directory contains the .graphml files for each of the genes of interest which can be opened in your network graphing software of choice. If you do not have a prefrence, we provide information about using Cytoscape below.

Visualization

scTIGER outputs .graphml files for each gene of interest as well as a overall interaction map. You can upload these files to your graphing software of choice. Additionally, we have included our Cytoscape style file (scTIGER_CytoscapeStyle.json) for your use.

The default style on Cytoscape does not show directionality and regulation type (pictured on the left). The scTIGER style automatically adds directionality arrows and displays regulation types as either a T for downregulation or an arrow for upregulation (pictured on the right). CytoscapeStyleImage

Information on importing the scTIGER style into Cytoscape can be found here

Paper

Corresponding Paper: Dautle M, Zhang S, Chen Y. scTIGER: A Deep-Learning Method for Inferring Gene Regulatory Networks from Case versus Control scRNA-seq Datasets. International Journal of Molecular Sciences. 2023; 24(17):13339. https://doi.org/10.3390/ijms241713339

Please cite this paper when using scTIGER.

Footnotes

  1. Heidegger, I., Fotakis, G., Offermann, A., Goveia, J., Daum, S., Salcher, S., Noureen, A., Timmer-Bosscha, H., Schäfer, G., Walenkamp, A., Perner, S., Beatovic, A., Moisse, M., Plattner, C., Krogsdam, A., Haybaeck, J., Sopper, S., Thaler, S., Keller, M. A., . . . Pircher, A. (2022). Comprehensive characterization of the prostate tumor microenvironment identifies CXCR4/CXCL12 crosstalk as a novel antiangiogenic therapeutic target in prostate cancer. Molecular Cancer, 21(1), 132. https://doi.org/10.1186/s12943-022-01597-7 ↩

  2. Chen, M. B., Jiang, X., Quake, S. R., & Südhof, T. C. (2020). Persistent transcriptional programmes are associated with remote memory. Nature (London), 587(7834), 437-442. https://doi.org/10.1038/s41586-020-2905-5 ↩