Zero-shot Benchmarking of RNA Language Models in Structural, Functional, and Evolutionary Learning
收藏官方服务:
资源简介:
Codes and datasets for RNA secondary structure prediction,zero-shot RNA classifications and zero-shot RNA fitness prediction. Compared to previous versions: (1) The number of evaluated RNA language models has increased, while the evaluation of protein language models has been removed; (2) For structural tasks, the original TS set was refined through family-level deduplication to create TS-Hard; (3) For classification tasks, the Rfam dataset underwent sequence sampling (ensuring intra-family sequence similarity below 80%) to generate RfamSample, while the ArchiveII dataset was deduplicated by RNA type to produce ArchiveII-Nr; (4) A fitness prediction task was added, utilizing the DMS assays from RNAgym dataset.
提供机构:
Zenodo创建时间:
2024-12-13



