ImmuDef

Project: A deep representation learning method for quantitative immune defense function evaluation and its clinical applications

Requirements

Python Package Requirements

• Python 3.10
• scikit-learn 1.2.2
• numpy 1.23.5
• pandas 1.5.3
• pytorch 2.0.1

R Package Requirements

• R 4.3.2
• getopt 1.20.4
• tidyverse 2.0.0
• GSVA 1.50.1
• clusterProfiler 4.10.1
• msigdbr 7.5.1

🧬 ssGSEA Immune Pathway Enrichment Analysis (C7 Gene Sets)

This R script performs single-sample Gene Set Enrichment Analysis (ssGSEA) using the MSigDB C7 immunologic signature gene sets. It takes a gene expression matrix as input and outputs enrichment scores for each sample across all C7 pathways.

---

📦 Required R Packages

Make sure the following packages are installed:

⚠️ Note: This script only supports human (Homo sapiens) data due to the use of org.Hs.eg.db.

---

📁 Input Format

✅ Requirements

• File type: CSV (comma-separated)
• Rows: Genes (using official HGNC gene symbols, e.g., TP53, CD3D)
• Columns: Samples
• Values: Expression values
  • Use raw counts (e.g., from RNA-seq) → script uses kcdf="Poisson"

Usage

Example (input.csv):

Rscript Rscript_ssGSEA.R -i input_file.csv -o output_file.csv

Beta-VAE Training Script

This script trains a Beta-VAE model on single-cell RNA sequencing data. Below are the instructions for running the script.

Prerequisites

Make sure you have the following dependencies installed:

pip install torch numpy scikit-learn json argparse

Training Model

python train_model_paras_combine.py

DImmuScore calculation

This script implements the complete computational workflow for deriving the DImmuScore based on QImmuDef-VAE latent representations and ensemble-model–trained reference samples.

Command-Line Usage

python run_immudef.py \
  --input data/expression_matrix.csv \
  --vae_model models/QImmuDefVAE.pt \
  --ensemble_model models/ensemble.pkl \
  --gene_list data/immune_gene_list.txt \
  --outdir results/

Required Input Files

Argument Description

--input	Expression matrix (.csv or .tsv). Rows = samples,
--vae_model	Pretrained QImmuDef-VAE weights (.pt).
--ensemble_model	Trained ensemble classifier combining SVM, XGBoost and Logistic Regression (.pkl).
--gene_list	List of immune-related genes used during VAE training.
--outdir	Output folder where all results will be saved.