multimolecule/ipknot_plus_plus-mafft
收藏资源简介:
IPknot++是一个RNA二级结构预测的基准数据集,随IPknot++论文发布,主要用于外部评估而非训练。它包含三个变体:ipknot_plus_plus(覆盖bpRNA-1m和Rfam 14.5目标的单序列结构,无对齐)、ipknot_plus_plus-ref(Rfam 14.5目标带Rfam参考对齐,用于共同结构预测)和ipknot_plus_plus-mafft(Rfam 14.5目标带MAFFT对齐,用于共同结构预测)。数据集提供id、序列和二级结构(点括号表示)等列,对齐变体还包含aligned_ids和aligned_sequences列。常见用途是在SPOT-RNA风格分割上训练模型后,用IPknot++评估泛化能力。
IPknot++ is a benchmark dataset for RNA secondary structure prediction, published concurrently with the IPknot++ research paper, and is designed exclusively for external model evaluation rather than training. It includes three variants: 1. ipknot_plus_plus: single-sequence structure prediction tasks covering bpRNA-1m and Rfam 14.5 datasets, with no sequence alignments provided; 2. ipknot_plus_plus-ref: tasks sourced from Rfam 14.5, paired with Rfam reference alignments for consensus secondary structure prediction; 3. ipknot_plus_plus-mafft: tasks sourced from Rfam 14.5, paired with MAFFT-generated alignments for consensus secondary structure prediction. The dataset provides core columns including id, sequence, and secondary structure (represented via dot-bracket notation), while the alignment-enabled variants additionally contain aligned_ids and aligned_sequences columns. A common use case is to evaluate the generalization performance of models trained on a SPOT-RNA-style dataset split using the IPknot++ benchmark.




