Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Qibitz: Mining PubMed for Repurposable Drugs

David Massart, Marc Zeicher

TL;DR

This work tackles the difficulty of systematically identifying repurposable drugs for specific pathologies via PubMed searches. It introduces Qibitz, a faceted search interface powered by two new specialized indexes: a drug field derived from a Belgic BelMed vocabulary and a gene field built on HGNC symbols, enriched through MeSH mappings, registry data, and unstructured-text extraction with LLM-based disambiguation. An ETL pipeline (extract, transform, load) automates data ingestion from PubMed into MongoDB and Elasticsearch, enabling a front-end faceted search that surfaces drug–gene and drug–pathology associations. The authors demonstrate the approach with three case studies (mucosal melanoma, Parkinson disease, and mucoepidermoid carcinoma), showing how Qibitz can reveal repurposed add-on therapies and new candidate drugs, thereby accelerating hypothesis generation for clinical investigation. Overall, the paper provides a practical, data-driven workflow for pattern discovery in literature metadata to support drug repurposing and collaboration between researchers and clinicians.

Abstract

PubMed's current search interface makes it tedious to systematically search for medical and research literature on drugs that could potentially be used to treat a given pathology, including patients with genetically altered tumors. This is because physicians must search separately for each drug-pathology combination (or drug-gene combination). To streamline this process, this paper proposes adding a faceted search interface to PubMed. Faceted search is a common feature on e-commerce websites that allows users to filter search results by selecting different fields. By incorporating this technology, not only can physicians save time and improve the accuracy of their literature searches, but also presenting search results in this way makes patterns emerge, which can suggest new treatment options for a given pathology (including patients with genetically altered tumors).

Qibitz: Mining PubMed for Repurposable Drugs

TL;DR

This work tackles the difficulty of systematically identifying repurposable drugs for specific pathologies via PubMed searches. It introduces Qibitz, a faceted search interface powered by two new specialized indexes: a drug field derived from a Belgic BelMed vocabulary and a gene field built on HGNC symbols, enriched through MeSH mappings, registry data, and unstructured-text extraction with LLM-based disambiguation. An ETL pipeline (extract, transform, load) automates data ingestion from PubMed into MongoDB and Elasticsearch, enabling a front-end faceted search that surfaces drug–gene and drug–pathology associations. The authors demonstrate the approach with three case studies (mucosal melanoma, Parkinson disease, and mucoepidermoid carcinoma), showing how Qibitz can reveal repurposed add-on therapies and new candidate drugs, thereby accelerating hypothesis generation for clinical investigation. Overall, the paper provides a practical, data-driven workflow for pattern discovery in literature metadata to support drug repurposing and collaboration between researchers and clinicians.

Abstract

PubMed's current search interface makes it tedious to systematically search for medical and research literature on drugs that could potentially be used to treat a given pathology, including patients with genetically altered tumors. This is because physicians must search separately for each drug-pathology combination (or drug-gene combination). To streamline this process, this paper proposes adding a faceted search interface to PubMed. Faceted search is a common feature on e-commerce websites that allows users to filter search results by selecting different fields. By incorporating this technology, not only can physicians save time and improve the accuracy of their literature searches, but also presenting search results in this way makes patterns emerge, which can suggest new treatment options for a given pathology (including patients with genetically altered tumors).

Paper Structure

This paper contains 25 sections, 1 figure, 1 table.

Table of Contents

  1. A Motivational Example
  2. Specializing Fields for Improved Filtering Capabilities
  3. Deriving a Drug Field
  4. Deriving a Gene Field
  5. Populating the Gene Field by Mapping MeSH Descriptors
  6. Populating the Gene Field by Mapping Registry Numbers
  7. Extracting HGNC Symbols from Unstructured Text Fields
  8. PubMed Data Acquisition and Transformation Workflow
  9. Extraction
  10. Transformation
  11. Importation
  12. Identifying Drugs as Potential Therapies for a Given Indication Using Qibitz: Three Examples
  13. Potential Repurposed Drugs Add-on Therapies Identified by Qibitz Based on Tumor Genes Alterations Revealed by FoundationOne Next-Generation Sequencing
  14. Mutations and their significance:
  15. Therapeutic implications
  16. ...and 10 more sections

Figures (1)

  • Figure 1: PubMed data acquisition and transformation workflow.