RegVar is a deep neural network-based computational server for prioritizing tissue-specific regulatory impact of human noncoding SNPs on their potential target genes. RegVar integrates the sequential, epigenetic and evolutionary conservation profiles of SNPs and their potential target genes in 17 human tissues, and give tissue-specific predictions of regulatory probabilities of the provided SNPs on provided genes.
A text file containing a list of SNP IDs and their possible target genes is required to be uploaded to the server for batch analysis.
The result will be generated based on all pairwise combinations of SNPs and genes. SNPs and genes lacking annotations are excluded and pairs of SNPs and genes that are located on different chromosomes are removed. The remaining pairs are referred to as valid pairs and RegVar would accept no more than 10,000 valid pairs.
Click here to see an example file
SNP ID(s) (indels are currently not supported) and their possible target gene(s) are accepted as input in the SNP and Gene search box, respectively. Multiple SNP IDs or genes should be delimited by commas, spaces or tabs, and if so, the result will be generated based on all pairwise combinations of SNPs and genes.
All results will be listed in the result page, including the basic information of your query data (the positions of the input SNPs and TSSs of genes and the genomic distance between them, in GRCh37/hg19 genome coordinates) (positions of TSSs are annotated from GTEx eGene list, v7 release), and the regulatory probabilities calculated by RegVar.
Raw probability scores come straight from the tissue-specific model, and are interpretable as the extent to which the SNP is likely to have an effect on the regulation of the corresponding gene in your selected tissue.
A result file containing the same information will be sent to your email address, if you have it input.
The RegVar website computes RegVar scores based on the DHS-filtered models trained on GTEx datasets. Besides, We also provide the scripts to train non-DHS-filtered models (or full models) on GTEx datasets and to train pathogenic RegVar models on HGMD dataset. Click the following download link for more information.
The datasets and source code to run RegVar locally are freely available at the download page.