VariBench is based on datasets of experimentally validated variations, against which variants can be compared to make predictions. The datasets of VariBench are compiled based on published literature and other public databases. The datasets are categorized in four sections:
1. Variants that affect protein tolerance.
2. Variants that affect protein stability.
3. Variants that affect transcription binding sites.
4. Variants that affects splicing sites (this section is actually very limited, as it containes information on just a couple of genes).
VariBench can also map variants to the sequences contained in RefSeq and within the 3D protein structures at Protein Data Bank (PDB).
