Scholarlink

<p>Breast cancer treatment and prognosis depend on accurate classification of hormone receptors (HR), particularly estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) enrichment status. Using copy number alteration (CNA or predictor gene) data, model predicts ER, PR, HER2 statuses, and classify breast cancer gene association. A new breast cancer classification model uses the Multi-Instance Learning with Deep Multimodal Attention Network (M-DMAN) to analyze gene (CNA) and biomarker data for breast cancer association. To estimate receptor enrichment status for each sample, the model uses gene expression profiles and biomarkers like ER, PR, and HER2. Predictions are used to classify breast cancer-associated samples. The M-DMAN model blends multimodal data with an attention mechanism that highlights the most informative aspects to fuse gene and biomarker data for accurate prediction. The model achieved 96.23% classification accuracy after extensive labelled data training. Statistics confirm the model’s stability and reliability in identifying breast cancer-associated samples. Multimodality and attention-based feature selection improve interpretability, making the model ideal for complicated medical datasets with essential gene and biomarker interactions. M-DMAN’s biomarker-driven breast cancer prediction capability makes it a promising tool for oncology clinical diagnostics and decision-making.</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

Gene-biomarker data-driven deep learning model with m-dman for er, pr, her2 enrichment and breast cancer association prediction

  • 期刊论文

  • 发表日期：

    2026年1月14日

查看全文
  • N. Banupriya,
  • T. Sethukarasi

摘要

Breast cancer treatment and prognosis depend on accurate classification of hormone receptors (HR), particularly estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) enrichment status. Using copy number alteration (CNA or predictor gene) data, model predicts ER, PR, HER2 statuses, and classify breast cancer gene association. A new breast cancer classification model uses the Multi-Instance Learning with Deep Multimodal Attention Network (M-DMAN) to analyze gene (CNA) and biomarker data for breast cancer association. To estimate receptor enrichment status for each sample, the model uses gene expression profiles and biomarkers like ER, PR, and HER2. Predictions are used to classify breast cancer-associated samples. The M-DMAN model blends multimodal data with an attention mechanism that highlights the most informative aspects to fuse gene and biomarker data for accurate prediction. The model achieved 96.23% classification accuracy after extensive labelled data training. Statistics confirm the model’s stability and reliability in identifying breast cancer-associated samples. Multimodality and attention-based feature selection improve interpretability, making the model ideal for complicated medical datasets with essential gene and biomarker interactions. M-DMAN’s biomarker-driven breast cancer prediction capability makes it a promising tool for oncology clinical diagnostics and decision-making.