Nomenclature-based data retrieval without prior annotation: facilitating biomedical data integration with fast...

Area of Science:

• Biomedical Informatics
• Computational Biology
• Data Science

Background:

• Assigning nomenclature codes to biomedical data is complex, costly, and prone to errors.
• Accurate data annotation is crucial for retrieving and analyzing medical and molecular biology information.
• Existing methods require extensive pre-coding, which can become obsolete with vocabulary updates.

Purpose of the Study:

• To introduce a new approach for retrieving biomedical data records without prior coding.
• To demonstrate a method that achieves efficient data retrieval through synonym matching.
• To eliminate the substantial effort and expense associated with traditional data annotation.

Main Methods:

• Utilized a recently described fast algorithm, the doublet method, for iterative searches.
• Searched a 105+ Megabyte corpus of PubMed abstracts using query terms matched against multiple vocabularies.
• Expanded query terms into arrays of equivalent search items, including over one hundred synonyms per term.

Main Results:

• Achieved fast query response times: 0.05 seconds with Snomed and 5 seconds with the Developmental Lineage Classification of Neoplasms.
• Demonstrated the doublet method's ability to search any vocabulary version without pre-annotation.
• Successfully retrieved records containing synonymous terms, matching the results of pre-coded searches.

Conclusions:

• The doublet method offers an efficient and cost-effective alternative to traditional biomedical data annotation.
• This approach enhances data discoverability and usability by accommodating any vocabulary version on-the-fly.
• Open-source Perl scripts are available for implementing and testing the doublet method algorithm.