Hi! This repository contains the code for our paper “Gene2role: A Role-Based Gene Embedding Method for Comparative Analysis of Signed Gene Regulatory Networks.” You can read the paper on BioRxiv: https://www.biorxiv.org/content/10.1101/2024.05.18.594807v1.abstract.
The repository is organized into two main sections:
1. Tutorial for Gene2role
2. Reproduction of the Paper’s Results
- pipeline.py
- tools/
- codes/
- eeisp.py
- merge_edgelist.py
- spearman.py
- split_cells.py
python pipeline.py TaskMode CellType EmbeddingMode input [**]
TaskMode. 1: run SignedS2V for an edgelist file.
2: run spearman and SignedS2V from gene X cell count matrix.
3: run eeisp and SignedS2V from gene X cell count matrix.
CellType. 1: single cell-type.
2: multiple cell-type.
EmbeddingMode. 1: single network embedding.
2: multiple network embedding, only work if the previous argument is 2.
input Input file, either a gene X cell matrix for TaskMode 2 and 3, or edgelist for TaskMode 1.
Other arguments, check "python pipeline.py --help"
python pipeline.py 3 2 1 data/singleCell/test_rna/count.csv --project TEST321 --OPT1 --OPT2 --OPT3 --workers 61 --until-layer 1 --threCDI 0.0001 --threEEI 0.0001 --cell_metadata data/singleCell/test_rna/metadata.csv
If you have edgelist for a single GRN and want to generate embedding representations, use the following basic command:
python pipeline.py 1 1 1 your_edgelist
If you have a gene X cell count matrix, want to construct a GRN , and generate embedding representations for a single cell type, use the following basic command:
EEISP (recommended):
python pipeline.py 3 1 1 your_count_matrix --threCDI 0.5 --threEEI 0.5
Spearman:
python pipeline.py 2 1 1 your_count_matrix --correlation_threshold 0.4
If you have two or more edgelists for GRNs and want to generate embedding representations simultaneously, use the following basic command:
python pipeline.py 1 2 2 your_edgelist1 your_edgelist2 your_edgelistx
If you have a gene X cell count matrix, want to construct GRNs, and generate embedding representations for multiple cell types, use the following basic command:
EEISP (recommended):
python pipeline.py 3 2 2 your_count_matrix --cell_metadata your_cell_metadata --threCDI 0.5 --threEEI 0.5
Spearman:
python pipeline.py 2 2 2 your_count_matrix --cell_metadata your_cell_metadata --correlation_threshold 0.4
This command will also generate an index tracker file (index_tracker.tsv) to clarify the geneIDs across multiple GRNs. Columns are cell types and rows are genes. The corresponding values show the replaced ID in the GRN of the gene in the specified cell type.
Replace your_edgelist, your_count_matrix, your_cell_metadata, and other placeholders with the actual file names or paths corresponding to your data.
your_edgelist is a TSV file representing the GRN with three columns, which are geneID1, geneID2, and edge sign (1/-1). The number of rows indicates the number of edges in the GRN.
your_count_matrix is a CSV file representing the Gene X Cell count matrix. Columns are cells, and rows are genes. The value indicates the gene expression level.
your_cell_metadata is a CSV file showing the relationship between cells and cell types. Columns are information about cells, and rows are cells. Two columns are required: [orig.ident] and [celltype]. [orig.ident] indicates the cell ID and must be aligned with the column names in your_count_matrix. [celltype] indicates the cell type of the cell.
All the processed data generated in the paper can be downloaded from https://figshare.com/articles/dataset/data/25852915.
The basic information about the data we used is as follows:
We used Gene2role to generate embedding by the following hyperparameters using the pipeline.py mention above.
- One simulated and four curated networks.
- Hyperparameters: 1 1 1
- Single-cell RNA-seq network generated from B cell in human PBMC dataset
- EEISP hyperparameters: 3 1 1
- Spearman Hyperparameters: 2 1 1
- Single-cell multi-omics network generated from Ery_0 stat in multi-omics networks.
- Hyperparameters: 1 1 1
- Human glioblastoma dataset.
- Hyperparameters: 3 2 2
- CD4 cells from PBMC dataset.
- Hyperparameters: 3 2 2
- Human BMMC dataset.
- Hyperparameters: 3 2 2
- Human PBMC dataset.
- Hyperparameters: 3 2 2
- single-cell multi-omics dataset.
- Hyperparameters: 1 2 2
The codes for clustering genes using lovain algorithm can be found in URL. We used the embeddings generated previously to analyze the gene module stability. Specifically, for human glioblastoma dataset, we used the embedding generated in Result 2. And for single-cell multi-omics dataset, we used the embedding generated in Result 3
The codes for downstream analyzed of the embeddings can be found in reproduce accordingly.
Feel free to reach Xin Zeng (wstxinzeng@gmail.com) and Shu Liu (Shu.liu.eq@gmail.com)to request files and more details from the analysis process!
This project is licensed under the MIT License.
MIT License
Copyright (c) 2024 Shu Liu
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.