Scholarlink

<p>Canine mammary tumors (CMT) are highly heterogeneous, and outcome varies greatly. Currently, survival prediction of CMT is mainly based on histopathological examination and Tumor-Node-Metastasis (TNM) staging. In human breast cancer, molecular markers such as estrogen receptor protein expression and gene expression signatures are important predictors of tumor progression and outcome. In dogs, however, neither of these are regularly used in clinical decision making. This study aimed to investigate whether gene expression biomarkers could improve patient stratification and survival prediction, thereby advancing precision medicine for dogs with mammary tumors.</p><p>We used a publicly available dataset of 146 CMTs including gene expression data and clinical variables. Using a machine learning approach, we fitted and compared three Cox survival models with different data types (clinical, gene expression and combined). Each model’s ability to discriminate risks was evaluated with Uno’s C-index and ROC-AUC at three timepoints. We found that age, ER status and tumor invasion to lymphatic vessels were essential for survival prediction. The two models that included gene expression were less accurate in discriminating&#xa0;between low- and high-risk dogs than the model including clinical variables only and did not provide additional value to clinical data. Finally, using the clinical model, we predicted patient risk which allowed us to stratify the cohort into two distinct groups, revealing significant differences in the expression of Hallmark signatures. Pathways related to proliferation and metabolism were enriched in high-risk patients, while immune-pathways were enriched in the low-risk group. In conclusion, ER status evaluated by immunohistochemistry and lymphatic invasion evaluated by tumor histology could be valuable additions to current TNM-staging. The role of gene expression in survival prediction of CMT remains unclear.</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

Unraveling prognostic factors in canine mammary gland tumors using machine learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月2日

查看全文
  • Hedda Fjell Scheel,
  • John Zobolas,
  • Tonje Gulbrandsen Lien,
  • Frode Lingaas,
  • Helga Bergholtz

摘要

Canine mammary tumors (CMT) are highly heterogeneous, and outcome varies greatly. Currently, survival prediction of CMT is mainly based on histopathological examination and Tumor-Node-Metastasis (TNM) staging. In human breast cancer, molecular markers such as estrogen receptor protein expression and gene expression signatures are important predictors of tumor progression and outcome. In dogs, however, neither of these are regularly used in clinical decision making. This study aimed to investigate whether gene expression biomarkers could improve patient stratification and survival prediction, thereby advancing precision medicine for dogs with mammary tumors.

We used a publicly available dataset of 146 CMTs including gene expression data and clinical variables. Using a machine learning approach, we fitted and compared three Cox survival models with different data types (clinical, gene expression and combined). Each model’s ability to discriminate risks was evaluated with Uno’s C-index and ROC-AUC at three timepoints. We found that age, ER status and tumor invasion to lymphatic vessels were essential for survival prediction. The two models that included gene expression were less accurate in discriminating between low- and high-risk dogs than the model including clinical variables only and did not provide additional value to clinical data. Finally, using the clinical model, we predicted patient risk which allowed us to stratify the cohort into two distinct groups, revealing significant differences in the expression of Hallmark signatures. Pathways related to proliferation and metabolism were enriched in high-risk patients, while immune-pathways were enriched in the low-risk group. In conclusion, ER status evaluated by immunohistochemistry and lymphatic invasion evaluated by tumor histology could be valuable additions to current TNM-staging. The role of gene expression in survival prediction of CMT remains unclear.