The variable quality of metadata about biological samples used in biomedical experiments

Area of Science:

• Bioinformatics
• Data Science
• Biomedical Research

Background:

• High-quality metadata is crucial for effective data management and analysis in biomedical research.
• Existing sample metadata repositories, such as NCBI's BioSample and EBI's BioSamples, are widely used.
• Ensuring the accuracy and standardization of metadata is essential for data reproducibility and integration.

Purpose of the Study:

• To analytically assess the quality of sample metadata in the BioSample (NCBI) and BioSamples (EBI) databases.
• To identify and quantify anomalies in metadata population and standardization.
• To evaluate the impact of metadata quality on data searchability and secondary use.

Main Methods:

• Analysis of 11.4 million sample metadata records from BioSample and BioSamples.
• Examination of metadata field names and values for standardization and adherence to requirements.
• Clustering of metadata field names to identify variations in representing sample aspects.

Main Results:

• Significant anomalies were found in metadata quality across both repositories.
• Metadata field names and values frequently lack standardization and controlled vocabularies.
• Inconsistent data types were observed even for simple binary or numeric fields.
• Multiple representations for the same sample attribute indicate a lack of standardization.

Conclusions:

• A lack of robust mechanisms for enforcing and validating metadata requirements was identified.
• The observed metadata deficiencies are likely to hinder effective data searching and secondary utilization.
• Improvements in metadata quality control are necessary to enhance the value of biomedical datasets.