On expert curation and scalability: UniProtKB/Swiss-Prot as a case study

Area of Science:

• Biomedical Informatics
• Bioinformatics
• Scientific Literature Analysis

Background:

• Biological knowledgebases, such as UniProtKB/Swiss-Prot, are vital for scientific research, providing curated knowledge.
• The increasing volume of biomedical literature challenges the capacity of knowledgebases to keep pace with new discoveries.

Purpose of the Study:

• To assess the literature triage process for UniProtKB/Swiss-Prot.
• To determine the ratio of relevant papers for curation within the growing biomedical literature.

Main Methods:

• Utilized the PubTator text-mining tool to tag over 10,000 articles for relevance assessment.
• Analyzed the number of papers evaluated versus curated by expert curators.
• Calculated the percentage of PubMed-indexed papers relevant for UniProt curation annually.

Main Results:

• UniProt curators evaluate 50,000-70,000 papers annually, curating 8,000-10,000.
• Approximately 90% of papers in PubMed fall outside the scope of UniProt.
• Only 2-3% of annually indexed PubMed papers are relevant for UniProt curation.

Conclusions:

• Measuring only curated publications underestimates the curation workload.
• Expert curation in UniProt is scalable despite the vastness of biomedical literature.
• Efficient literature triage is crucial for maintaining comprehensive biological knowledgebases.