Variation benchmark datasets: update, criteria, quality and applications.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Developing and testing computational methods requires experimental data for performance assessment.
• Benchmark datasets with verified outcomes are essential for evaluating new algorithms.
• Existing databases like VariBench and VariSNP facilitate sharing of variation benchmark datasets.

Purpose of the Study:

• To update and expand the VariBench database with new datasets.
• To facilitate the development and benchmarking of computational methods for genetic variation analysis.
• To enable comparison of new method performances against existing studies.

Main Methods:

• Incorporation of 419 new datasets from 109 papers into VariBench.
• Categorization of datasets into 20 groups based on variant type and effect.
• Description of variants at DNA, RNA, and protein levels, including structural information.

Main Results:

• The updated VariBench contains over 329 million variants, reducing redundancy.
• Datasets cover diverse variation types, including insertions, deletions, substitutions, and effects on DNA, RNA, and protein properties.
• Performance comparison of the PON-P2 predictor demonstrated the utility of benchmark studies.

Conclusions:

• The updated VariBench database enhances the development and testing of new computational methods for variation interpretation.
• Facilitates standardized performance comparison of prediction tools.
• Highlights the need for detailed information and data sharing to overcome limitations in comparative studies.