Cross-Species Splicing Variant Benchmark Datasets (Human Data1&2, Rat, Pig, Chicken)

Benchmark datasets for evaluating splicing variant prediction models in the manuscript Comprehensive review and assessment of multi-species splicing variant prediction: task-specific deep learning models and genomic foundation models. This release provides a dataset package for benchmarking covering five datasets: two human benchmarks (human Data1 and human Data2) and three non-human benchmarks from rat, pig, and chicken. Each dataset contains balanced positive (splice-altering) and negative (normal) variants with default 8,192 bp sequence windows. Files included: splicevardb_benchmark.tsv: 23,877 human variants (human Data1, SpliceVarDB, GRCh38) humansQTL_benchmark. tsv: 22,948 human variants (human Data2, GRCh38) ratsQTL_benchmark.tsv: 28,120 rat variants (Data3, Rnor_6.0) pigsQTL_benchmark.tsv: 26,358 pig variants (Data4, Sscrofa11.1) chickensQTL_benchmark.tsv: 25,000 chicken variants (Data5, GRCg6a) This upload was organized as a cleaned benchmark release for the revised manuscript, with unified file packaging across all five datasets. Related GitHub repository for pre-processing: https://github.com/ByronSyun/DL-CrossSpeciesReview.git