Machine learning classifiers for smooth muscle cell differentiation prediction
Year: 2026
Type: software
Abstract: Script Brief purpose script_a_repeated_cv_fast-4.py Performs repeated stratified 5-fold cross-validation for gene-expression-space XGBoost and Random Forest classifiers. It evaluates two predefined binary classification tasks and outputs fold-level and summary AUC, Cohen’s kappa, and accuracy statistics with empirical 95% intervals. script_a_repeated_cv_fast-4.py script_b_gene_stability_compact-3.py Performs repeated stratified resampling to evaluate the stability of gene rankings for XGBoost and Random Forest models. It varies feature-selection settings, derives XGBoost SHAP-based and Random-Forest importance-based gene lists, and exports per-fold lists, selection frequencies, and stability summaries for the original 15-gene signature. script_b_gene_stability_compact-3.py extract_extended_genes.py Extracts additional stable candidate genes from the aggregated gene-frequency output of the stability analysis, excluding the original 15-gene signature. It applies predefined stability thresholds to generate separate Random Forest and XGBoost candidate-gene lists. extract_extended_genes.py run_3ml-5.py Implements an extended repeated-cross-validation framework comparing XGBoost, Random Forest, neural-network, and PCA–XGBoost models across the two binary classification tasks. It calculates AUC, Cohen’s kappa, and accuracy and includes SHAP-based feature-ranking procedures. run_3ml-5.py comment5_total_100-2.py A companion extended repeated-cross-validation and sensitivity-analysis script using 100 repeated 5-fold resampling iterations. It compares XGBoost, Random Forest, neural-network, and PCA–XGBoost approaches and generates detailed performance summaries and model-derived gene rankings. comment5_total_100-2.py pc_xgb-6.py Conducts a PCA–XGBoost benchmark analysis. It standardizes the gene-expression matrix, retains principal components explaining 95% of variance, trains XGBoost classifiers on the reduced feature space, and reports AUC, Cohen’s kappa, and the number of retained components. pc_xgb-6.py
Author: lin, chorng horng
Institution: Dayeh University
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7213599400