A general computational framework for associating cancer cell-intrinsic driver genes with -extrinsic cell-cell communication
- Driver2Comm is now available for any non-tumor cell type of interest in the TME. (May 27th, 2024)
The following software must be installed on your machine:
Python: tested with version 3.10
R : tested with R 4.0.3
- numpy (tested v1.26.2)
- pandas (tested v2.1.4)
- statsmodels (tested v0.14.1)
- scipy (tested v1.11.4)
- networkx (tested v3.2.1)
- matplotlib (tested v)
- CytoTalk(tested v0.99.0)
Create a new conda environment using the environment.yml file or the requirements.txt file with one of the following commands:
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/bioconda conda env create -f environment.yml
Install R and the Cytotalk package
conda install R R
> install.packages("devtools") >devtools::install_github("tanlabcode/CytoTalk")
Install Driver2Comm install Driver2Comm using pip
pip install Driver2Comm -U
- For detail information of Cytotalk installation, please refer to tanlabcode/Cytotalk
the input of Driver2Comm include two part:
- intrinsic factors: the driver information of patients in this study
- Extrinsic factors: the MCTC networks of patients in this study
Driver2Comm needs patient driver information as input，where column1 is patients' id, the remaining columns are driver involving in your study. 1 means the driver variated in this patient. 0 means not.
| Patient_id | Driver1 | Driver2 | |
|---|---|---|---|
| 0 | patient1 | 1 | 0 |
| 1 | patient2 | 0 | 0 |
| 2 | patient3 | 1 | 0 |
| 3 | patient41 | 1 | 1 |
If you use CytoTalk as your MCTC network construction algorthm, the input data folder should be agreed with following structure.
── CytoTalk_output
├─ patient1
├─ Celltype1-Celltype2
├─ FinalNetwork.txt
├─ Celltype1-Celltype3
├─ FinalNetwork.txt
└─ Celltype2-Celltype3
├─ FinalNetwork.txt
├─ patient2
├─ Celltype1-Celltype2
├─ FinalNetwork.txt
├─ Celltype1-Celltype3
├─ FinalNetwork.txt
└─ Celltype2-Celltype3
├─ FinalNetwork.txt
├─ patient3
├─ ...
└─ patient4
├─ ...
fromDriver2Commimportformulate_c2c_networkinputPATH='./data/CytoTalk_output'outputPATH='./data/Driver2Comm_input'patient_list= ['patient1','patient2',...,'patientn']
formulate_c2c_network(inputPATH,outputPATH,patient_list)Warning !!!1. the order of patient_list should exactly coincide with patient driver information matrix
Then in your output directory, you will find a c2c_network.data , and the c2c_network.data is in following format.
Driver2Comm can extend to other CCC inferences method, in this case, you can also directly formulate you cell-cell communication as the following format.
t # 1 # t # 1 means the MCTC network of patient 1
v 0 FERMT3__Macrophages # vertex data was formualed as 'v vertex_id vertex_label'
v 1 MITF__Macrophages
e 0 1 0 # edge was formualed as 'e vertex_id_a vertex_id_b edge label'
...
t # n
v 1 SOCS1__Cd8+Tcells
v 1 CD52__Cd8+Tcells
e 0 1 0
t # -1 # t # -1 represent as the end of this file2. the label of celltype should not contain any space for example: "CD8+ T" × "CD8+_T" ✓
You can apply Driver2Comm in the following step.
- Step0: Construction of cell-cell communication network
- Step1: Identification of cancer-driver assoication communicaiton patterns
- Step2: Downstream analysis of Driver2Com
Rscript construction_ccc_network.RAfter construction of CCC network by Cytotalk, we started to identify the cancer-driver associated communicaiton patterns. The rest analysis of Driver2Comm is fully based on python implement. Based on the excellent flexibility and interactivity of jupyter notebook, we recommend you to use notebook for all of Driver2Comm's remaining analyses.
importDriver2Commasdcc2c_network=dc.read_c2c_network(os.path.join('./data/c2c_network.data'))
model=dc.Driver2Comm(
c2c_network=c2c_network,
patient_driver_info=patient_driver_info,
minsup=20,
association_test_threshold=0.05,
outputPATH='./data/Driver2Comm_output',
cancer_type='Brca'""" parameters: __________________ :param c2c_network: the formulated cell-cell communication network data :param patient_driver_info: the patient metadata formulated as the form shown in previous section "intrinsic factor" :param minsup: Hyperparameter, the minimal suppport of frequent subnetwork extraction algorithm :param association_test_threshold: Hyperparameter, the threshold of association testing :param outputPATH: the path where output files place :param cancer_type: type of cancer __________________ """
)
# run Driver2Comm to identify cancer driver associated CCC signatures.model.run()We offer several ways to visualize and evaluate the cancer driver associated CCC signatures identified by Driver2Comm.
# display all cancer driver associated CCC signatures on screen.model.display_associated_FP()
# output association testing result of Driver2Comm as a Pandas Dataframeassociated_mat=model.output_assocaited_mat()we provide two function to get IE pathway
importDriver2Commasdcinterested_driver='ESR1__Tumor'# your interested driverpatient_path='./xxx'# the path containing cytotalk result of interested patientk=10# number of interested signaturestop_k_ie_pathway_mat=dc.generate_top_k_ie_pathway(model.association_test_ret,model.frequent_subgraphs,interested_driver, k=k,patient_path=patient_path)importDriver2Commasdcinterested_driver='ESR1__Tumor'# your interested driverpatient_path='./xxx'# the path containing cytotalk result of interested patientinterested_gene='APOE_Macrophages'# your interested genepcsf_network, original_network=dc.read_graph_from_cytotalk_output(patient_path)
ie_ret=dc.get_ie_pathway(interested_driver,interested_gene,pcsf_network,original_network)Runzhi Xie (rzxie@stu.xidian.edu.cn)
The gSpan implementation in Driver2Comm was based in part on source code of gspan-mining package.