ThermoScan

ThermoScan extracts thermodynamic and protein stability data from full-text articles in PubMed Central's Open Access subset to support prediction of amino acid variant effects on protein structure and function.

Key Features:

• Text mining with regular expressions: Uses regular expressions to identify groups of words including key conceptual terms, thermodynamic variables, and their units within full-text articles.
• Empirical scoring system: Computes an empirical score for each article to rank the likelihood that it reports thermodynamic data on protein stability.
• Optimization and validation: Extraction and scoring procedures were optimized using the ProTherm dataset and validated on a manually curated set of articles containing thermodynamic data.
• Full-text performance advantage: Comparative analyses indicate ThermoScan outperforms text-mining algorithms focused on abstracts by improving prediction accuracy and data retrieval from full-text sources.

Scientific Applications:

• Predictive modeling of variant effects: Aggregates thermodynamic measurements to support models that predict the impact of amino acid substitutions on protein function and stability.
• Structural biology: Provides experimental thermodynamic data to inform analyses of protein stability and structure–function relationships.
• Molecular genetics and bioinformatics: Supports dataset curation and computational analyses related to mutagenesis and variant interpretation.

Methodology:

Scans full-text articles for specific textual patterns indicative of protein stability experiments, applies regular expressions to detect conceptual terms, thermodynamic variables, and units, and computes an empirical score per article; the approach was optimized using the ProTherm dataset and validated on a manually curated article set.