Fig. 1. The U1 sequence as a key feature for predicting RNA nuclear retention.
a, Overview of the experimental design. b, Western blots showing the protein expression of lamin-B and β-tubulin in nuclear and cytoplasmic fractions from human/macaque brain tissue. c, Distribution of normalized N/C ratios of mRNAs and lncRNAs in brain tissue. n = 15,734 mRNAs and 1,861 lncRNAs for human; n = 15,180 mRNAs and 2,719 lncRNAs for macaque; two-sided, unpaired Wilcoxon test, P < 2.2 × 10−16 and P < 2.2 × 10−16, respectively. The boxes represent interquartile range, with the line across the box indicates the median. The whiskers extend to the lowest and the highest values in the dataset. ***P ≤ 0.001. d, Sequences of transcripts excessively distributed in the nucleus or cytoplasm in HEK293T cells. e, Deep learning classification model to investigate the key cis elements underpinning the varied transcript nuclear export activity. f, Evaluation metrics, as well as features of prediction for the classification model, are shown. g, Key cis elements differentiating transcripts with varied nuclear export activity were identified by prioritizing the activation values in the first convolutional layer of this CNN network.