Related Experiment Video
Updated: Jul 12, 2025
07:35
A Knowledge Graph Approach to Elucidate the Role of Organellar Pathways in Disease via Biomedical Reports
Published on: October 13, 2023
1.7K
Serial KinderMiner (SKiM) discovers and annotates biomedical knowledge using co-occurrence and transformer models
Robert J Millikin1, Kalpana Raja1,2, John Steill1
1Morgridge Institute for Research, Madison, WI, USA.
BMC Bioinformatics
|November 2, 2023
Summary
Serial KinderMiner (SKiM) is a new literature-based discovery tool that helps researchers find hidden connections between biomedical concepts. This open-source algorithm and web interface identify relationship types and support large-scale queries for drug repurposing and disease research.
Area of Science:
- Biomedical Informatics
- Computational Biology
- Data Mining
Background:
- The exponential growth of biomedical literature necessitates efficient tools for knowledge discovery.
- Literature-based discovery (LBD) aims to uncover novel associations between concepts in disparate research domains.
- Existing LBD tools often lack flexibility, interpretability, or scalability for complex biomedical research queries.
Purpose of the Study:
- To introduce Serial KinderMiner (SKiM), an open-source LBD algorithm and web interface.
- To address limitations of current LBD tools, including relationship type identification, user-defined term lists, and scalability for large concept sets.
- To facilitate the discovery of A-B-C relationships between biomedical concepts.
Main Methods:
- Developed the Serial KinderMiner (SKiM) algorithm for identifying statistically significant A-B-C linkages.
- Integrated SKiM with a knowledge graph utilizing transformer machine-learning models for relationship interpretation.
- Created an open-source, user-friendly web interface with comprehensive biomedical concept lists.
Main Results:
- Demonstrated SKiM's efficacy in discovering A-B-C linkages through experiments in classic LBD, drug repurposing, and cancer-related research.
- Successfully supplemented SKiM with a knowledge graph to provide relationship type labels.
- Launched an accessible web interface (skim.morgridge.org) for performing SKiM searches.
Conclusions:
- SKiM is a versatile, domain-general algorithm for LBD searches.
- The tool enables discovery of relationships between user-defined concepts, supporting thousands of C terms.
- SKiM enhances LBD by identifying relationship types, moving beyond simple association detection.
