MarkerHub

MarkerHub identifies and ranks candidate biomarkers for hepatocellular carcinoma (HCC) by mining PubMed literature and applying machine learning methods to prioritize candidates for diagnostic, prognostic, and therapeutic research.

Key Features:

• Automated Literature Mining: Uses biomedical text mining to extract relevant information from published literature.
• Hybrid System Integration: Employs a hybrid approach combining machine learning–based and pattern‑based methods and integrates PubMed E-Utilities to collect abstracts.
• Named Entity Recognition (NER): Applies NER to identify genes and proteins mentioned in evidential sentences within the literature.
• Convolutional Neural Network Classification: Classifies identified genes and proteins as candidate biomarkers using a convolutional neural network (CNN).
• Ranking and Curation: Ranks extracted biomarkers using criteria including keyword frequency, article count, and journal impact factor.
• Comprehensive Database: Compiles a curated list of 2,128 candidate biomarkers extracted from PubMed publications spanning 2008 to 2017.

Scientific Applications:

• Biomarker Prioritization: Narrows reported biomarker candidates to a prioritized list for follow-up studies.
• Clinical Validation Support: Provides prioritized candidates to facilitate clinical validation efforts for HCC biomarkers.
• HCC Research Enablement: Supports diagnostic, prognostic, and therapeutic research in hepatocellular carcinoma by supplying literature‑derived candidate biomarkers.

Methodology:

Collects abstracts via PubMed E-Utilities; applies biomedical text mining with a hybrid machine learning–based and pattern‑based system; uses named entity recognition to extract genes and proteins from evidential sentences; classifies candidates with a convolutional neural network; ranks biomarkers by keyword frequency, article count, and journal impact factor; and compiles the resulting list of 2,128 candidates from PubMed 2008–2017.