Scholarlink

<p>Detection of cancer biomarker is a crucial step in early diagnosis and targeted therapy. This paper presents a filter–wrapper computational pipeline that leverages adaptive β-hill climbing (ABHC)-aided metaheuristic optimization algorithms for identification of biomarkers and detection of cancers from microarray gene expression. The integration of ABHC with metaheuristics enhances the search efficiency and accuracy in identifying underlying biomarkers. In the filtering stage, combined scores of mutual information and minimum redundancy maximum relevancy are utilized to select the top <i>p</i>% genes from high-dimensional microarray datasets. Subsequently, in the wrapper stage the candidate genes are passed to ten different ABHC-aided metaheuristics to reveal the most informative genes marking cancers using a KNN classifier. Experimental findings reflect that among the ten hybrid methods, ABHC-aided whale optimization has outperformed with an average of 100% accuracy with only 2.85 genes across seven datasets. Furthermore, comparative analysis implied that the obtained results have surpassed many state-of-the-art models in terms of maximum accuracy with minimum number of marker genes.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Adaptive \(\beta \)-hill climbing enhanced metaheuristic approach for cancer identification from gene expression

  • 期刊论文

  • 发表日期：

    2025年4月9日

查看全文
  • Emon Asad,
  • Ayatullah Faruk Mollah

摘要

Detection of cancer biomarker is a crucial step in early diagnosis and targeted therapy. This paper presents a filter–wrapper computational pipeline that leverages adaptive β-hill climbing (ABHC)-aided metaheuristic optimization algorithms for identification of biomarkers and detection of cancers from microarray gene expression. The integration of ABHC with metaheuristics enhances the search efficiency and accuracy in identifying underlying biomarkers. In the filtering stage, combined scores of mutual information and minimum redundancy maximum relevancy are utilized to select the top p% genes from high-dimensional microarray datasets. Subsequently, in the wrapper stage the candidate genes are passed to ten different ABHC-aided metaheuristics to reveal the most informative genes marking cancers using a KNN classifier. Experimental findings reflect that among the ten hybrid methods, ABHC-aided whale optimization has outperformed with an average of 100% accuracy with only 2.85 genes across seven datasets. Furthermore, comparative analysis implied that the obtained results have surpassed many state-of-the-art models in terms of maximum accuracy with minimum number of marker genes.