Machine learning analysis to identify key inflammatory genes in psoriasis using RNA-seq data and validate biosensor targets.

This project uses RNA-seq data from psoriatic and healthy skin samples to:

• Build a classification model to distinguish inflamed vs. normal tissue
• Identify the most predictive genes using SHAP analysis
• Validate NF-κB pathway genes as biomarkers for inflammation
• Support biosensor development targeting NO and H₂O₂

Source: GSE54456 from NCBI GEO

• Samples: 174 skin biopsies (92 psoriatic, 82 normal)
• Type: Bulk RNA-seq expression data (RPKM values)
• File: GSE54456_RPKM_samples.txt

pip install pandas numpy scikit-learn matplotlib seaborn shap

1. Download the expression data file GSE54456_RPKM_samples.txt from GEO
2. Place it in the same directory as the script
3. Run the analysis:

python psoriasis_analysis.py

• Loads gene expression data for 174 samples
• Labels samples as inflamed (psoriatic) or normal based on metadata
• Filters out low-expression genes (mean RPKM > 1.0)
• Applies log2 transformation for normalization

• Selects top 500 most variable genes
• Force-includes NF-κB pathway genes (NFKB1, NFKB2, RELA, etc.)
• Creates feature matrix for machine learning

• Splits data: 70% training, 30% testing
• Trains logistic regression with L2 regularization (C=0.1)
• Evaluates performance with accuracy and ROC-AUC

• Calculates SHAP values to identify most predictive genes
• Ranks genes by their contribution to inflammation prediction
• Highlights NF-κB pathway genes

The analysis outputs:

• Model Performance: Training/testing accuracy, ROC-AUC score
• Top 20 Predictive Genes: Ranked by SHAP importance
• NF-κB Pathway Analysis: Specific evaluation of inflammation regulators
• Biological Interpretation: Connection to NO/H₂O₂ biosensor targets

• NF-κB is a master regulator of inflammation
• In psoriasis, activated NF-κB drives inflammatory gene expression
• NF-κB activation leads to production of NO and H₂O₂

This analysis validates that:

• NO production: NF-κB activates iNOS enzyme → produces nitric oxide
• H₂O₂ production: NF-κB activates NADPH oxidase → produces hydrogen peroxide

Your plasmid circuit detecting NO and H₂O₂ targets molecules downstream of the key inflammatory genes identified here.

The script prints detailed results including:

• Data loading and preprocessing statistics
• Model performance metrics
• Classification report
• Top predictive genes with SHAP scores
• NF-κB pathway gene rankings
• Biological interpretation

• The model uses regularization (C=0.1) to reduce overfitting
• Sample labels are manually curated from GEO metadata
• SHAP provides interpretable feature importance
• NF-κB genes are force-included regardless of variance

• Dataset: Swindell et al., GSE54456, NCBI GEO
• SHAP: Lundberg & Lee (2017), "A Unified Approach to Interpreting Model Predictions"
• Biological context: NF-κB pathway in psoriasis pathogenesis

Open source for research and educational purposes.