tipit_benchmark_RNA

This repository provides a Python implementation of several transcriptomic signatures that were associated with immunotherapy response in the literature, for different cancer types and checkpoint inhibitors.

It contains the code used in our study to perform a benchmark of transcriptomic signatures to predict immunotherapy outcome in non-small cell lung cancer:

"Integration of clinical, pathological, radiological, and transcriptomic data improves the prediction of first-line immunotherapy outcome in metastatic non-small cell lung cancer"

Preprint: https://doi.org/10.1101/2024.06.27.24309583

Note: The transcriptomic signatures were selected based on the work of Kang et al. 2023.

Installation

Dependencies

• gseapy (=1.1.3)
• pandas (= 1.5.3)
• pyyaml (>= 6.0)
• scikit-learn (>= 1.2.0)

Optional (to run the scripts):

• scikit-survival (>= 0.21.0)
• tqdm (>= 4.63.0)
• xgboost (>= 1.7.5)

Install from source

Clone the repository:

git clone https://github.com/sysbio-curie/tipit_benchmark_RNA

Examples

Define and compute a transcriptomic signature

import pandas as pd
from benchmark_RNA.signatures import get_CYT_score

data = pd.read_csv("data/transcritpomic_data.csv", index_col=0)

#1. Define the function score
CYT_fun = get_CYT_score(data)

#2. Compute the scores
CYT_scores = data.agg(CYT_fun, axis=1)

Note: data should be a pandas DataFrame with samples in rows and genes in columns. Columns names should be gene symbols.

Some signatures include training pre-processing steps in their definition such as PCA (e.g., FTBRS, TME), scaling (e.g., TIS, IIS), or KNN (MFP). It may be required to define them and compute their values with different datasets.

Define and compute a transcriptomic signature with train and test data

import pandas as pd
from benchmark_RNA.signatures_gsea import get_MFP_score

data_train = pd.read_csv("data/transcritpomic_data_train.csv", index_col=0)
data_test = pd.read_csv("data/transcritpomic_data_test.csv", index_col=0)

#1. Define the function score
MFP_fun = get_MFP_score(data_train)

#2. Compute the scores
MFP_scores = data_test.agg(MFP_fun, axis=1)

Scripts

We provide a Python script to reproduce the benchmark of transcriptomic signatures for the prediction of immunotherapy outcome in lung cancer in our paper. It defines and tests the different signatures across the fold of a repeated cross-validation scheme.

python extract_signatures.py -c config.yaml

Available transcriptomic signatures

Name	Signature type	Cancer type	Immmune Checkpoint	References
CRMA	Marker genes	Melanoma	CTLA-4	Shukla et al.
CTLA4	Marker genes	Multiple	PD-L1	Herbst et al.
CX3CL1	Marker genes	Multiple	PD-L1	Herbst et al.
CXCL9	Marker genes	Melanoma	PD-L1	Qu et al.
CYT	Marker genes	Multiple	PD-1, CTLA-4	Rooney et al.
EIGS	Marker genes	Multiple	PD-1	Ayers et al.
ESCS	Marker genes	Urothelial cancer	PD-1	Wang et al.
FTBRS	Marker genes	Multiple	PD-L1	Mariathasan et al.
HLADRA	Marker genes	Melanoma	PD-1, PD-L1	Johnson et al.
HRH1	Marker genes	Multiple	PD-1, PD-L1, CTLA-4	Li et al.
IFNgamma	Marker genes	Multiple	PD-1	Ayers et al.
Immunopheno	Marker genes	Multiple	PD-1, CTLA-4	Charoentong et al.
IMPRES	Marker genes	Melanoma	PD-1, CTLA-4	Auslander et al.
IRG	Marker genes	Cervical cancer	PD-1, PD-L1, CTLA-4	Yang et al.
MPS	Marker genes	Melanoma	PD-1, CTLA-4	Pérez-Guijarro et al.
PD1	Marker genes	Multiple	PD-1	Taube et al.
PDL1	Marker genes	Multiple	PD-1, PD-L1	Herbst et al.
PDL2	Marker genes	Multiple	PD-1	Yearley et al.
Renal101	Marker genes	Renal cell carcinoma	PD-1, PD-L1	Motzer et al.
TIG	Marker genes	Multiple	PD-1	Cristescu et al.
TLS	Marker genes	Melanoma	PD-1, CTLA-4	Cabrita et al.
TME	Marker genes	Gastric cancer	PD-1, PD-L1, CTLA-4	Zeng et al.
APM	GSEA	Renal cell carcinoma	PD-1	Senbabaoglu et al.
CECMdown	GSEA	Multiple	PD-1	Chakravarthy et al.
CECMup	GSEA	Multiple	PD-1	Chakravarthy et al.
IIS	GSEA	Renal cell carcinoma	PD-1	Senbabaoglu et al.
IMS	GSEA	Gastric cancer	PD-1, PD-L1	Lin et al.
IPRES	GSEA	Multiple	PD-1	Hugo et al.
MFP	GSEA	Multiple	PD-1, PD-L1, CTLA-4	Bagaev et al.
MIAS	GSEA	Melanoma	PD-1	Wu et al.
PASSPRE	GSEA	Melanoma	PD-1	Du et al.
TIS	GSEA	Renal cell carcinoma	PD-1	Senbabaoglu et al.
CD8T_CIBERSORT	Deconvolution	Multiple	PD-1	Tumeh et al.
CD8T_MCPcounter	Deconvolution	Multiple	PD-1	Tumeh et al.
CD8T_Xcell	Deconvolution	Multiple	PD-1	Tumeh et al.
Immuno_CIBERSORT	Deconvolution	Melanoma	PD-1	Nie et al.

Acknowledgements

This repository was created as part of the PhD project of Nicolas Captier in the Computational Systems Biology of Cancer group and the Laboratory of Translational Imaging in Oncology (LITO) of Institut Curie.