Scholarlink

Background <p>Reliable tumor origin assessment from routinely archived specimens remains challenging, particularly for diagnostically ambiguous liver malignancies. Although formalin-fixed, paraffin-embedded (FFPE) tissues are widely available, deep and quantitative site-specific N-glycoproteomics of these specimens is limited by formaldehyde-induced crosslinking and incomplete peptide recovery.</p> Methods <p>We developed GlyFFPE, a quantitative platform for site-specific N-glycosylation profiling from FFPE archival tissues, and applied it to discovery and independent verification datasets from colorectal cancer liver metastasis (CRLM) and liver hepatocellular carcinoma (LIHC) patients. We then integrated glyco-signatures with clinicopathological variables using an ensemble machine-learning classifier and evaluated performance in a temporal validation cohort. Model robustness and incremental value were assessed using a pathology-only comparator, repeated nested cross-validation, feature-selection stability analysis, and a same-institution temporal cohort.</p> Results <p>GlyFFPE enabled deep site-specific N-glycoproteome profiling from FFPE tissues and revealed metastasis-associated glyco-signatures characterized by elevated levels of high-mannose glycans on cell-adhesion molecules, prominently CEACAM1 and CEACAM5. The two-glycopeptide panel from CEACAM high-mannose signature showed modest discrimination between CRLM and LIHC in an independent TMT-PRM verification cohort (AUC = 0.73), indicating the need of expanding the targeted panel. The model integrating glyco-signatures with clinicopathological variables substantially improved diagnostic performance, achieving glycoproteomic-assisted discrimination in a same-institution temporal cohort. Repeated nested cross-validation provided a more conservative performance estimate with a mean AUC of 0.92 and a mean AUPRC of 0.96. Feature-selection analysis identified repeatedly selected molecular features, although moderate feature-set variability remained because of the limited cohort size. In addition to classification, GlyFFPE revealed metastasis-associated glycosylation remodeling, including increased high-mannose occupancy on alpha-1-acid glycoprotein 1 (ORM1) and coordinated alterations in extracellular matrix–associated glycoproteins.</p> Conclusion <p>GlyFFPE establishes an FFPE-compatible workflow for site-specific N-glycosylation profiling and identifies candidate molecular signatures that complement routine pathological assessment for tumor origin evaluation in challenging liver malignancies. Given the modest cohort size and the same-institution temporal evaluation, multicenter, blinded external validation are required before clinical implementation.</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

Deep N-glycoproteomic profiling of FFPE tissues enables precise diagnosis of primary liver cancer from metastatic liver cancer

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年8月3日

查看全文
  • Hanjie Li,
  • Ruizhen Bai,
  • Chenyu Huang,
  • Xiuyuan Wang,
  • Zeyang Yu,
  • Biao Zhou,
  • Ye Yuan,
  • Quan Liu,
  • Ganglong Yang,
  • Xiao-Dong Gao

摘要

Background

Reliable tumor origin assessment from routinely archived specimens remains challenging, particularly for diagnostically ambiguous liver malignancies. Although formalin-fixed, paraffin-embedded (FFPE) tissues are widely available, deep and quantitative site-specific N-glycoproteomics of these specimens is limited by formaldehyde-induced crosslinking and incomplete peptide recovery.

Methods

We developed GlyFFPE, a quantitative platform for site-specific N-glycosylation profiling from FFPE archival tissues, and applied it to discovery and independent verification datasets from colorectal cancer liver metastasis (CRLM) and liver hepatocellular carcinoma (LIHC) patients. We then integrated glyco-signatures with clinicopathological variables using an ensemble machine-learning classifier and evaluated performance in a temporal validation cohort. Model robustness and incremental value were assessed using a pathology-only comparator, repeated nested cross-validation, feature-selection stability analysis, and a same-institution temporal cohort.

Results

GlyFFPE enabled deep site-specific N-glycoproteome profiling from FFPE tissues and revealed metastasis-associated glyco-signatures characterized by elevated levels of high-mannose glycans on cell-adhesion molecules, prominently CEACAM1 and CEACAM5. The two-glycopeptide panel from CEACAM high-mannose signature showed modest discrimination between CRLM and LIHC in an independent TMT-PRM verification cohort (AUC = 0.73), indicating the need of expanding the targeted panel. The model integrating glyco-signatures with clinicopathological variables substantially improved diagnostic performance, achieving glycoproteomic-assisted discrimination in a same-institution temporal cohort. Repeated nested cross-validation provided a more conservative performance estimate with a mean AUC of 0.92 and a mean AUPRC of 0.96. Feature-selection analysis identified repeatedly selected molecular features, although moderate feature-set variability remained because of the limited cohort size. In addition to classification, GlyFFPE revealed metastasis-associated glycosylation remodeling, including increased high-mannose occupancy on alpha-1-acid glycoprotein 1 (ORM1) and coordinated alterations in extracellular matrix–associated glycoproteins.

Conclusion

GlyFFPE establishes an FFPE-compatible workflow for site-specific N-glycosylation profiling and identifies candidate molecular signatures that complement routine pathological assessment for tumor origin evaluation in challenging liver malignancies. Given the modest cohort size and the same-institution temporal evaluation, multicenter, blinded external validation are required before clinical implementation.