NLProt

NLProt extracts protein names from natural language biomedical text, such as PubMed abstracts, to identify mentions and link them to corresponding sequence database entries.

Key Features:

• Protein name extraction: Extracts and identifies protein names from natural language biomedical text, including PubMed abstracts.
• Hybrid filtering: Applies dictionary- and rule-based filtering to candidate name mentions.
• Machine learning: Integrates multiple support vector machines (SVMs) for protein name tagging.
• Performance: Reports a precision of 75% and a recall of 76% when partially tagged names are counted as errors.
• Novel name detection: Detects protein names that are not present in predefined dictionaries.
• Ambiguity handling: Addresses frequent and ambiguous terms encountered in Drosophila studies (e.g., "white", "wing", "bizarre").
• Linking to sequence databases: Associates extracted protein names with corresponding entries in sequence databases.
• Supported inputs: Handles PubMed/MEDLINE identifiers, author names, titles, journals, collections of abstracts, and entire papers.

Scientific Applications:

• Literature curation: Supports construction of custom datasets and databases derived from scientific literature.
• Database annotation: Enhances annotation of biological sequence databases by providing literature-derived name mappings.
• Automated text mining: Enables large-scale extraction of protein mentions from PubMed abstracts and other biomedical texts.
• Drosophila-specific analysis: Improves identification of protein mentions in contexts with frequent ambiguous common terms.

Methodology:

Combines dictionary- and rule-based filtering with multiple support vector machines (SVMs) to tag protein names in biomedical text.