m42-health/variant-benchmark

Variant Benchmark是一个基因组学基准数据集，设计用于评估模型在多种生物学背景下利用变体信息的效果。它包含多个子数据集，分别针对编码和非编码致病变异评估、基因表达效应预测、选择性剪接、DNA甲基化影响评估、祖先分类以及常见与合成变异的区分等任务。数据集以变体信息为核心，包含了大量的基因组变异和相关的生物信息学特征。

Variant Benchmark is a genomics benchmark dataset designed to evaluate how effectively models leverage variant information across diverse biological contexts. It consists of multiple sub-datasets, each targeting tasks such as coding and noncoding pathogenicity assessment, gene expression effect prediction, alternative splicing, DNA methylation effect assessment, ancestry classification, and distinguishing between common and synthetic variants. The dataset is centered around variant information and includes a large number of genomic variations and associated biological features.