SORTA: a system for ontology-based re-coding and technical annotation of biomedical phenotype data.

Area of Science:

• Biomedical Informatics
• Data Science
• Biobanking

Background:

• Standardizing biomedical data semantics, such as phenotypes, is essential for comparative and integrative analyses.
• Retrospective data standardization is often necessary due to varied data collection protocols.
• Manual data curation for matching to ontologies like SNOMED CT, ICD-10, and HPO is time-consuming.

Purpose of the Study:

• To develop SORTA, a computer-aided system to mechanize the encoding of free text or locally coded biomedical data to formal coding systems or ontologies.
• To evaluate SORTA's applicability and efficiency in real-world use cases.

Main Methods:

• SORTA matches original data values to target coding systems (Excel, OWL, OBO formats).
• It employs Lucene and n-gram based matching algorithms for semi-automatic candidate code shortlisting.
• The system can learn from human expert matches and handles semicolon-delimited data input.

Main Results:

• In the LifeLines biobank, SORTA recoded 90,000 free text values to Metabolic Equivalent of Task (MET) codes with 0.97 precision and 0.98 recall.
• In the CINEAS system, SORTA mapped to Human Phenotype Ontology (HPO), achieving 0.58 precision and 0.45 recall.
• Users reported SORTA as a significant time saver and quality improvement, reducing human error.

Conclusions:

• SORTA effectively automates and streamlines the data (re)coding process.
• The system demonstrates significant potential to aid numerous projects requiring biomedical data standardization.
• SORTA enhances data quality and accelerates research by reducing manual curation efforts.