Benchmark results.

A comparison of three secondary structure prediction algorithms, using shape data from Deigan et al. [15] for the three RNA molecules, yeast aspartyl tRNA (asp-tRNA), hepatitis C virus internal ribosomal entry site (HCV IRES), and the P546 domain from the bI3 group I intron (P546), along with shape data from [26] for three additional RNA molecules, E. coli phenylalanine tRNA (phe-tRNA), E. coli 5S ribosomal RNA (5S rRNA), and F. nucleatum glycine riboswitch (glycine). The benchmark results are tabulated for (A) RNAsc+shape, (B) RNAstructure+shape, and (C) RNAstructure (with no shape data). Sensitivity is abbreviated by sens., positive predictive value is abbreviated by ppv. The average pointwise entropy, Morgan-Higgs structural diversity, and the expected distance of the computed probabilities to the probing data are abreviated by ave ent., str. div., and edist., respectively. Not shown: results for medloop and V. vulnificus adenine riboswitch (1Y26), for which all three methods have optima sensitivity and ppv values of 1.0.

本研究针对三种RNA二级结构预测算法开展对比实验，采用Deigan等[15]报道的三类RNA分子的形状数据（shape data），分别为酵母天冬氨酸转运RNA（yeast aspartyl tRNA, asp-tRNA）、丙型肝炎病毒内部核糖体进入位点（hepatitis C virus internal ribosomal entry site, HCV IRES）以及bI3组I型内含子的P546结构域（P546 domain from the bI3 group I intron, P546）；同时使用文献[26]提供的另外三种RNA分子的形状数据，包括大肠杆菌苯丙氨酸转运RNA（E. coli phenylalanine tRNA, phe-tRNA）、大肠杆菌5S核糖体RNA（E. coli 5S ribosomal RNA, 5S rRNA）以及具核梭杆菌甘氨酸核糖开关（F. nucleatum glycine riboswitch）。基准测试结果以表格形式汇总，涵盖（A）RNAsc+shape、（B）RNAstructure+shape以及（C）未使用形状数据的RNAstructure三种方案。灵敏度（Sensitivity）缩写为sens.，阳性预测值（positive predictive value, PPV）缩写为ppv。平均逐点熵、Morgan-Higgs结构多样性以及计算所得概率与探测数据的期望距离，分别缩写为ave ent.、str. div.与edist.。未展示的结果包括medloop与创伤弧菌腺嘌呤核糖开关（V. vulnificus adenine riboswitch, 1Y26），三类算法在该两类样本上的最优灵敏度与阳性预测值均为1.0。