A benchmark database for variations

Home | Instructions | Datasets | Citing | Disclaimer |


VariBench is a benchmark database suite comprising of validated variation datasets collected from literature which can be applied to systematically analyse the performance of computational variant effect predictors.  These benchmark datasets can also be used for testing and training novel predictors in this field. Currently, VariBench houses datasets for testing and training four different variations affecting – (a) protein tolerance (b) protein stability (c) transcription factor binding sites (d) splice sites. Additionally, the database provides residue-residue level mapping of the variant positions to:

The benchmark variation datasets in VariBench are experimentally validated for their effects and are designed to make them reusable as benchmarks. Overlapping entries, if present, in the training datasets used by the method developers and the benchmark datasets can be filtered out with the help of database identifiers and the variation information provided in VariBench. VariBench save the time of the researchers as it is laborious to generate benchmarks. The datasets and their variant position mappings are given in the download section.