KiPar

KiPar retrieves textual documents containing kinetic parameters for metabolic pathway modeling in yeast systems biology as a standalone Java application.

Key Features:

• Modular multi-reaction search: Allows specification of multiple reactions and their associated kinetic parameters within a single search request.
• Identifier-based retrieval: Accepts database identifiers such as EC numbers, GO, and SBO identifiers to reference and retrieve concept-specific documents.
• Integrative text mining: Leverages public data and software resources to develop large-scale text mining capabilities for biological information retrieval.
• Concept-to-synonym indexing: Maps identified concepts to their synonyms within textual documents to facilitate precise search queries.
• Performance metrics: Reports precision of 60% for abstracts and 48% for full-text articles versus baseline 44% and 24%, with recall enhanced by 36% for abstracts and doubled for full-text documents.
• Full-text extraction and novelty: Extracts more comprehensive kinetic data from full-text articles and, when combining abstracts and full text, achieves relative recall of 88% and novelty ratio of 92%.
• Validation against Boolean searches: Demonstrates superior performance compared to traditional Boolean search baselines.

Scientific Applications:

• Genome-scale metabolic model parameterization: Retrieves experimental kinetic parameters required for constructing detailed genome-scale metabolic models.
• Yeast systems biology: Targets kinetic parameter extraction relevant to yeast metabolic pathway modeling.
• Kinetic data curation and expansion: Supports identification of new and relevant documents beyond existing curated literature to expand kinetic parameter collections.

Methodology:

Uses identifier-based input (EC numbers, GO, SBO), large-scale text mining leveraging public data and software resources, indexing that maps concepts to synonyms, a modular multi-reaction search approach, and validation by comparison to Boolean searches with reported precision and recall metrics.