TRACEvar: Prioritizing and interpreting pathogenic variants in tissue contexts

These datasets were used to train the TRACEvar machine-learning scheme in 17 human tissues. The datasets contain the features and labels of 67,933 genetic variants for two human genome versions (hg37, hg38). Per dataset, each variant was associated with 1,313 features, including 116 variant-specific features extracted from the CADD V1.6 and 1,197 tissue-specific features extracted from the TRACE machine-learning scheme [https://www.biorxiv.org/content/10.1101/2021.02.16.430825v1], https://netbio.bgu.ac.il/trace/. Each variant was labeled in 29 tissues according to whether it leads to a disease that manifests in that tissue (i.e., pathogenic) or not (i.e., benign). TRACEvar is available at https://netbio.bgu.ac.il/tracevar, for source code please visit https://github.com/ChananArgov/TRACEvar.