Fig. 5: Inter-family secondary structure prediction for RiNALMo and homology-based tools.
From: RiNALMo: general-purpose RNA language models can generalize well on structure prediction tasks
a Secondary structure prediction average F1 scores for the ArchiveII evaluation datasets. The numbers in brackets next to each RNA family name represent the count of RNAs for which at least one homolog was identified, followed by the total number of RNAs in that dataset. RNAs for which no homologs were found are ignored. The best result for each evaluation dataset in the tables is shown in bold. b Distribution of secondary structure prediction F1 scores for different tools on the ArchiveII evaluation datasets (sample sizes n: 1278, 738, 510, 456, 442, 37, and 35, respectively). Box plots show the median (center line), 25th and 75th percentiles (bounds of box), whiskers extending to the smallest and largest values within 1.5× the interquartile range, and individual outliers beyond the whiskers. Minimum and maximum values are 0 and 1, respectively.
