Implementation of an open chemistry knowledge base with a Semantic Wiki

Charlotte Neidiger, Tarek Saier, Kai Kühn, Victor Larignon, Michael Färber, Claudia Bizzarri, Helena Simek Tosino, Laura Holzhauer, Michael Erdmann, An Nguyen, Dean Harvey, Pierre Tremouilhac, Claudia Kramer, Daniel Hansch, Fabian Schönle, Jana Alpin, Maximilian Hartmann, Jérome Wagner, Nicole Jung, Stefan BräseJ. Cheminformatics 2025 · DOI

TL;DR

This paper introduces the Chemistry Knowledge Base (CKB), an open-access platform built on Semantic MediaWiki, designed to collaboratively collect, structure, and share chemical research data. By integrating cheminformatics tools and standardized data forms, CKB aims to overcome the limitations of traditional publication models and commercial databases, making chemical knowledge more accessible and usable for the scientific community.

What Problem Does This Paper Address?

The current landscape of scientific information dissemination, particularly in chemistry, faces several challenges. Traditional journal publications, while offering peer review, often present results in unstructured, non-machine-readable formats, hindering efficient data reuse and analysis. While data repositories and pre-print servers are emerging, a comprehensive and collaborative platform for structured chemical knowledge is lacking. Commercial databases like SciFinder and Reaxys, though valuable, are expensive and can have a time lag in incorporating new findings. This creates a gap for scientists seeking to access, compare, and build upon existing research efficiently, especially with the rise of AI-driven research methods that thrive on structured data.

Methods

The core of this work is the development of an open chemistry knowledge base (CKB) using Semantic MediaWiki (SMW). SMW was chosen for its user-friendly interface, robust versioning capabilities, and its ability to store information in both unstructured (textual) and structured (machine-readable) formats.

Key methodological contributions include:

  • Enhanced SMW for Chemistry: SMW was extended to specifically support the collection and structured summarization of chemical data.
  • Standardized Data Forms and Model: Data forms were designed along with a data model to ensure standardized input and organization of research results. This includes mapping data from current research to original sources.
  • Cheminformatics Integration: Tools for capturing chemical structures in machine-readable formats were implemented. This involved integrating the Ketcher JavaScript editor for drawing molecules and ChemScanner for parsing molecular groups and generating identifiers like InChIKeys and SMILES.
  • Ontology-Based Data Modeling: External ontologies such as OBI, ChEBI, and MOP were utilized to define classes in the data model, ensuring semantic enrichment and coherent representation of knowledge across systems.
  • Hierarchical Content Structure: The CKB is organized into interconnected pages:
  • Topic pages: Provide an overview of scientific fields, summarizing research with text, tables, and molecular drawings.
  • Publication pages: Summarize the content of journal articles or datasets, including text, tables, and molecules, and can link to literature or other publications.
  • Investigation pages: Detail experimental data, categorized into "molecular processes" and "assays," serving as the granular data entries.
  • Molecule pages: Represent individual chemical structures as distinct objects, allowing for unique naming, referencing, and integration with external databases.

Datasets

The paper does not explicitly list datasets used for training or evaluation in the traditional machine learning sense. Instead, the CKB itself is built to ingest and organize data from scientific publications. The primary "dataset" is the collection of chemical research results that are entered into the CKB. The paper uses the example of "photocatalytic CO2 reduction" to demonstrate the platform's capabilities, implying that data from publications in this specific research area were used to populate and test the CKB. No new benchmark datasets were introduced.

Key Insights and Ideas

  • Bridging the Gap: The CKB effectively bridges the gap between unstructured scientific literature and the need for structured, machine-readable data, which is crucial for modern research, including AI applications.
  • Community-Driven Knowledge: By leveraging Semantic MediaWiki, the platform fosters a collaborative environment where scientists can contribute and curate knowledge, reducing reliance on commercial entities.
  • Semantic Enrichment: The use of ontologies and standardized data models allows for a deeper, semantically rich understanding of chemical information, enabling more sophisticated queries and data integration.
  • Reusability of Chemical Structures: Representing molecules as distinct objects (Molecule pages) that can be drawn once and reused across multiple contexts significantly enhances efficiency and consistency.
  • Open Science Paradigm: The CKB embodies the principles of open science by providing free access to a structured knowledge base, promoting transparency and collaboration.

Key Results

The paper demonstrates the successful implementation of the CKB, showcasing its ability to:

  • Integrate Chemical Data: Effectively integrate chemical research results into a collaboratively usable platform.
  • Structure Information: Provide structured and standardized information through the use of forms and a defined data model.
  • Visualize Structures: Allow for the drawing and representation of chemical structures in both human-readable (images) and machine-readable formats.
  • Facilitate Comparison: Enable effective data comparison and contextual analysis within the Wiki environment.
  • Demonstrate Use Case: Successfully demonstrated its potential by organizing and comparing research in "CO2 reduction in homogeneous photocatalytic systems."

While quantitative results like accuracy metrics are not presented, the qualitative success lies in the functional implementation of a robust, open, and collaborative knowledge base for chemistry.

Limitations and Future Work

Limitations:

  • Current Scope: The CKB is currently demonstrated with a specific example ("photocatalytic CO2 reduction") and is intended to be expanded to other fields of chemical science.
  • Manual Input: While some automation is envisioned, the current process still relies on manual input of data, which can be time-consuming.
  • SMW Limitations: The paper briefly mentions limitations of SMW (SI Sect. 11), though these are not detailed in the main text.

Future Work:

  • Large Language Models (LLMs): Integration of LLMs is a primary focus for transforming unstructured scientific data into structured formats, with an "assistant" mode for user review.
  • Natural Language Query Interface: Developing an interface that allows users to interact with the system using everyday language to lower the barrier to accessing complex data.
  • Semantic Description of Reactions: Full semantic description of chemical reactions.
  • Extension to Other Fields: Expanding the CKB to encompass other areas of chemical science.
  • Automated Checks and Curation: Further development of tools to support automatic checks and curation to identify and correct errors.

Summary updated

BibTeX
@article{neidiger2025implementation,
  author = {Charlotte Neidiger and Tarek Saier and Kai Kühn and Victor Larignon and Michael Färber and Claudia Bizzarri and Helena Simek Tosino and Laura Holzhauer and Michael Erdmann and An Nguyen and Dean Harvey and Pierre Tremouilhac and Claudia Kramer and Daniel Hansch and Fabian Schönle and Jana Alpin and Maximilian Hartmann and Jérome Wagner and Nicole Jung and Stefan Bräse},
  title = {{Implementation of an open chemistry knowledge base with a Semantic Wiki}},
  journal = {J. Cheminformatics},
  year = {2025},
  doi = {10.1186/s13321-025-01037-w},
}
Figure