VIST - a Variant-Information Search Tool for precision oncology

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Cancer diagnosis and treatment rely on analyzing patient genomic mutations.
• Existing biomedical search engines often prioritize basic science over clinically relevant findings.
• There is a need for specialized tools to access publications pertinent to clinical cancer care.

Purpose of the Study:

• To develop a search engine, the Variant-Information Search Tool (VIST), for identifying clinically relevant publications based on oncological mutation profiles.
• To address the limitations of general search engines in providing clinically actionable information for cancer treatment.

Main Methods:

• VIST indexes PubMed abstracts and ClinicalTrials.gov content.
• Advanced text mining identifies genes, variants, and drugs.
• Machine learning algorithms score publications for clinical relevance.
• A user-friendly web interface provides access to VIST's functionality.

Main Results:

• VIST demonstrates superior ranking of clinically relevant documents compared to PubMed and standard vector space models.
• The tool effectively targets publications with direct impact on cancer treatment decisions.

Conclusions:

• Standard search engines inadequately cater to diverse user needs in scientific literature retrieval.
• VIST offers a specialized solution for finding clinically relevant cancer research.
• The machine learning-based architecture of VIST can serve as a model for domain-specific search engines.