Recall and bias of retrieving gene expression microarray datasets through PubMed identifiers.

Area of Science:

• Bioinformatics
• Genomics
• Data Science

Background:

• Locating gene expression microarray datasets is crucial for resource reuse.
• Current database annotations are often insufficient for complex queries.
• This study evaluates PubMed identifiers for dataset discovery.

Purpose of the Study:

• To assess the efficacy of PubMed identifiers in finding gene expression microarray datasets.
• To compare dataset retrieval using PubMed IDs versus data sharing statements.
• To determine if PubMed ID searches yield representative datasets.

Main Methods:

• Conducted a scripted search of GEO and ArrayExpress using PubMed identifiers.
• Compared retrieved datasets with those found via data sharing statements.
• Analyzed dataset characteristics, including research theme, technology, and citation metrics.

Main Results:

• PubMed ID searches retrieved 76.6% of publicly available datasets, compared to 91.4% found via data sharing statements.
• Searches using PubMed IDs identified datasets with larger sample sizes and higher citation counts.
• Individual database searches (GEO/ArrayExpress) yielded lower retrieval rates.

Conclusions:

• PubMed identifiers identify a majority of datasets but may exclude studies from lower-impact journals.
• Authors should ensure complete citation details for dataset submissions.
• Accurate citation facilitates dataset visibility and contribution to future research.