ResidueFinder

ResidueFinder identifies mentions of individual amino acid residues and mutations in primary research literature using expanded regular expressions implemented in Python to enable high-recall extraction for database construction and analysis.

Key Features:

• Enhanced Regular Expression Library: Uses an expanded set of regular expressions to detect native amino acids and mutation mentions across diverse textual representations in full texts and abstracts, reducing false negatives compared with MutationFinder.
• Precision and Recall Optimization: Provides adjustable parameters to balance precision and recall and evaluates performance using expanded F-measure statistics (Fβ) to prioritize recall or precision via β.
• Efficiency Improvements ("cut" option): Offers a streamlined regular expression option named "cut" that increases processing speed by up to several hundredfold with a modest decrease in recall for large-scale text analyses.
• Artifact Handling: Incorporates specific regular expression elements to mitigate common artifacts from PDF-to-text conversion, improving identification accuracy in converted documents.
• Versatile Corpus Support: Validated on both full texts and abstracts and suitable for constructing databases of residue mentions across PubMed and other article collections.

Scientific Applications:

• Molecular biology residue extraction: Facilitates extraction of residue-level mentions from literature to support studies of specific amino acids within protein sequences.
• Protein function and structure analysis: Aggregates residue evidence from publications to inform investigations of protein function and structural impacts of mutations.
• Large-scale database construction and text mining: Enables creation of comprehensive databases of residue mentions across PubMed/full-text corpora for large-scale bioinformatics projects.

Methodology:

Implemented in Python and built upon MutationFinder, ResidueFinder applies an expanded regular expression library (including elements to handle PDF-to-text artifacts), offers a streamlined "cut" regex option for speed, exposes adjustable parameters for precision/recall tuning, and evaluates performance with F-measure statistics (Fβ).