OncoScore

OncoScore prioritizes cancer-associated genes by text-mining PubMed citation frequencies to rank genes for interpretation of mutations from Next Generation Sequencing (NGS) experiments.

Key Features:

• Text-mining approach: Employs text-mining of biomedical literature via dynamic web queries to PubMed to extract citation frequencies.
• Gene ranking: Ranks genes by association with cancer based on extracted citation frequencies, producing prioritized lists of candidate genes.
• NGS mutation prioritization: Applies ranking to mutation lists generated by Next Generation Sequencing (NGS) experiments to aid interpretation.
• Validation metrics: Validated using Receiver Operating Characteristic (ROC) curves and Area Under the Curve (AUC) on manually curated datasets, reporting an AUC of 90.3% (95% Confidence Interval: 88.1–92.5%).
• Cut-off threshold: Uses a cut-off threshold of 21.09 to flag genes with significant literature support as potentially oncogenic.

Scientific Applications:

• Cancer genomics: Prioritizes candidate cancer genes from large-scale mutation datasets to support cancer genomics studies.
• Target selection and hypothesis generation: Guides targeted investigations, hypothesis generation, and selection of genes for experimental follow-up and therapeutic research by ranking literature-supported associations.

Methodology:

OncoScore extracts and analyses citation frequencies from PubMed using text-mining and dynamic web queries, ranks genes by cancer association, and evaluates performance with ROC curves and AUC on manually curated datasets using a cut-off threshold of 21.09.