DeepGene: Dual ensemble explainable pipeline for gene-based enhanced biomarker selection
Year: 2025
Type: article
Abstract: Accurate and interpretable biomarker identification is crucial for effective cancer diagnosis and treatment. Conventional feature selection methods often struggle with instability and lack of transparency, especially when handling complex high-dimensional data. This paper presents the Dual Ensemble Explainable Pipeline for Gene-based Enhanced (DeepGene) Model to address these twin challenges. DeepGene integrates two ensemble strategies: one amalgamating filter-based techniques to discern the most discriminative and robust genes and another for identifying interpretable biomarkers through aggregating different explainable machine learning methodologies. A novel two-stage aggregation approach within the filter-based ensemble technique enhances gene selection, improving stability, exploration, and information integration. A comprehensive evaluation of public cancer microarray datasets demonstrates DeepGene's superior performance in stability and interpretability compared to state-of-the-art methods, resulting in a 3-10% improvement in classification accuracy. These findings highlight DeepGene's potential to significantly advance cancer research and clinical practice, ultimately benefiting patients and healthcare providers. • A two-stage aggregation enhances feature selection in gene expression data. • An ensemble SHAP method identifies relevant, interpretable genes. • DeepGene integrates filter and SHAP ensembles for stable biomarker gene selection. • DeepGene achieves 3–10% gains on six datasets for early cancer biomarker discovery.
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: gold
Estimated APC paid: $2,490
OpenAlex ID: W4415436525