Fig 1.
Study workflow for development and validation of the miRNA-based diagnostic model.
Data preprocessing, feature selection, machine learning model development, external validation, interpretability analysis, and functional enrichment were performed sequentially using independent training and validation cohorts.
Table 1.
Characteristics of Study Cohorts.
Fig 2.
Study Overview and Data Processing.
(A) Sample distribution across different cohorts and platforms. (B) Principal component analysis before and after batch correction demonstrating successful removal of platform-specific batch effects while preserving biological signal separation.
Fig 3.
Differential Expression Analysis.
(A) Volcano plots showing differentially expressed miRNAs across datasets. (B) Expression heatmap of top dysregulated miRNAs between cancer and control groups.
Table 2.
Top Differentially Expressed miRNAs (Expression-Level Analysis).
Table 3.
Machine Learning Model Performance Comparison.
Fig 4.
Machine Learning Model Performance.
(A) ROC curves comparing different algorithms. (B)Calibration plots showing predicted vs observed probabilities.
Fig 5.
SHAP summary plot showing feature importance rankings.