The Variant Effect Prediction (VEP) Web Application is designed to provide researchers with an intuitive and efficient way to explore the regulatory potential of noncoding single-nucleotide variants, making it a powerful resource for genomic analysis. This app predicts which noncoding variants are likely to impact chromatin accessibility and downstream gene regulation. This tool leverages ChromBPNet models, which have been pre-trained on different cell types, allowing users to predict the cell type-specific effects of different variants.
This tool is designed for variant effect prediction of common single-nucleotide variants. We have pre-computed results for more than 22 million single-nucleotide variants, representing all variants from gnomad v3.1.2 that are present in more than 1% of individuals from any population.
The model predictions are most reliable within peak regions of chromatin accessibility and so the user interface displays information on whether the given variant is located within a called peak region.
We estimate the effect of a variant using the Jensen-Shannon Divergence (JSD) which represents the difference in prediction between the reference and alternate alleles. For variants with large predicted effects, we provide the base-pair-resolution importance scores surrounding the variant. These show which bases of DNA are most important to the model and often correspond to transcription factor motifs.
All plots are available for download as high-resolution rasterized and vectorized images, and the underlying raw data can be exported for further downstream analysis.
For further information on how to interpret the various plots and outputs, we refer users to the User Guide.
Users can enter a variant using its genomic coordinates in the following format: chr#:pos:ref:alt (e.g., chr1:123456:A:T). This input requires specifying the chromosome number, the base pair position (hg38), the reference allele, and the alternate allele. This method allows for precise analysis of any SNP based on its genomic location.
Users can input the variant using its reference SNP ID (rsID), which is a standardized identifier from the dbSNP database (e.g., rs123456). This input provides a quick and convenient way to retrieve predictions for well-characterized variants without needing to specify genomic coordinates.