Skip to content

Latest commit

 

History

History
79 lines (47 loc) · 3.96 KB

README.md

File metadata and controls

79 lines (47 loc) · 3.96 KB
Raw

Example Usage for The scDrug

Preprocessing

The example data is composed of a random 10% subdata from GSE156625: Onco-fetal reprogramming of endothelial cells drives immunosuppressive macrophages in Hepatocellular Carcinoma (scRNA-seq).

Before going through the steps in the scDrug, download and uncompress the example data data.zip from figshare and put the files under the example folder.

Single-Cell Data Analysis

  • First, we execute Single-Cell Data Analysis on the 10x-Genomics-formatted mtx directory data/10x_mtx, with batch correction of PatientID in the metadata data/metadata.csv, and clustering at resolution 0.6. Additionally, we assign arguments --annotation and --gsea to perform cell type annotation and Gene Set Enrichment Analysis (GSEA).
  • Required memory: 2.5GB

Note: This step may take a few minutes.

mkdir write/clustering

python3 ../script/single_cell_analysis.py --input data/10x_mtx --output write/clustering \
--metadata data/metadata.csv --batch PatientID --resolution 0.6 \
--annotation --gsea --GEP False
  • Inspecting the preceding output stored in write/clustering/scanpyobj.h5ad, we regard the clusters with tumor cell percentages over twice the normal cell percentages, which consist of clusters 1, 5 and 9, as the tumor clusters. Then, we apply Single-Cell Data Analysis once again to carry out sub-clustering on the tumor clusters at automatically determined resolution.

Note: To accelerate the process of automatically determined resolution, increase the number of CPU with arguments --cpus CPUS.

mkdir write/subclustering

python3 ../script/single_cell_analysis.py --input write/clustering/scanpyobj.h5ad --output write/subclustering \
--format h5ad  --clusters '1,5,9' --auto-resolution --cpus 4

Drug Response Prediction

  • Based on the sub-clustering result write/subclustering/scanpyobj.h5ad, we run Drug Response Prediction to predict clusterwise IC50 and cell death percentages to drugs in the GDSC database.
  • Required memory: 2GB
mkdir write/drug_response_prediction

python3 ../script/drug_response_prediction.py --input write/subclustering/scanpyobj.h5ad --output write/drug_response_prediction

Treatment Selection

In Treatment Selection, we first impute cell fractions of bulk GEPs from the LINCS L1000 database with single-cell GEP write/subclustering/GEP.txt. Then, we selecte treatment combinations from the LINCS L1000 database with the CIBERSORTx result, and visualize the result treatment effect.

Impute Cell Fractions

  • Since it takes several hours to impute cell fractions, the result of CIBERSORTx/fractions, data/CIBERSORTx_Adjusted.txt and the L1000 instance info file data/GSE70138_Broad_LINCS_inst_info_2017-03-06.txt, is provided for the next step.
  • Required memory: 18GB

Note: With USERNAME and TOKEN acquired from CIBERSORTx, we could also run the following commands to impute cell fractions on previously generated write/subclustering/GEP.txt with celltype HEPG2 assigned. Notice that this could take several hours.

mkdir write/CIBERSORTx_fractions

python3 ../script/CIBERSORTx_fractions.py --input write/subclustering/GEP.txt --output write/CIBERSORTx_fractions \
--username USERNAME --token TOKEN --celltype HEPG2

Select Treatment Combinations

  • With the CIBERSORTx/fractions result data/CIBERSORTx_Adjusted.txt and the L1000 instance info file data/GSE70138_Broad_LINCS_inst_info_2017-03-06.txt as input, we select treatment combinations from the LINCS L1000 database with celltype HEPG2 assigned.
  • Required memory: <1GB
mkdir write/treatment_selection

python3 ../script/treatment_selection.py --input data/CIBERSORTx_Results.txt --output write/treatment_selection \
--celltype HEPG2 --metadata data/GSE70138_Broad_LINCS_inst_info_2017-03-06.txt