Related Experiment Video
Updated: Aug 6, 2026
09:45
Isolation and Profiling of Human Primary Mesenteric Arterial Endothelial Cells at the Transcriptome Level
Published on: March 14, 2022
CellExLink: End-to-end cell-type recognition and normalization in biomedical text
Alimire Nabijiang1, Leili Shahriyari1
1Department of Mathematics and Statistics, University of Massachusetts Amherst, Amherst, Massachusetts, United States of America.
Plos Computational Biology
|July 24, 2026
Summary
CellExLink automatically identifies and standardizes cell types in biomedical text, improving data consistency. This pipeline links diverse cell mentions to Cell Ontology (CL) identifiers for better research applications.
Area of Science:
- Biomedical Informatics
- Computational Biology
- Natural Language Processing
Background:
- Biomedical literature uses varied terminology for cell types, hindering automated analysis.
- Accurate cell-type recognition and normalization are crucial for data integration and knowledge discovery.
Purpose of the Study:
- To develop and evaluate CellExLink, an automated pipeline for cell-type mention recognition and normalization.
- To standardize cell-type mentions to Cell Ontology (CL) identifiers.
Main Methods:
- Developed an end-to-end pipeline, CellExLink, for cell-type recognition and normalization.
- Fine-tuned and evaluated the system on diverse biomedical corpora (articles, abstracts, captions).
- Assessed performance using exact and relaxed span F1 scores for recognition and normalization.
Main Results:
- CellExLink achieved macro-average F1 scores of 0.766 (exact span) and 0.855 (relaxed span) for cell mention recognition.
- CL identifier normalization achieved F1 scores ranging from 0.690 to 0.874.
- End-to-end evaluation yielded F1 scores up to 0.725, outperforming existing systems.
Conclusions:
- CellExLink effectively recognizes and normalizes cell-type mentions in biomedical text.
- The pipeline provides standardized Cell Ontology (CL) identifiers, facilitating downstream applications.
- This tool supports literature curation, relation extraction, and knowledge graph construction.
