Self-attention enabled deep learning of dihydrouridine (D) modification on mRNAs unveiled a distinct sequence signature from tRNAs

Dihydrouridine (D) is a modified pyrimidine nucleotide universally found in viral, prokaryotic, and eukaryotic species. It serves as a metabolic modulator for various pathological conditions, and its elevated levels in tumors are associated with a series of cancers. Precise identification of D sites on RNA is vital for understanding its biological function. A number of computational approaches have been developed for predicting D sites on tRNAs; however, none have considered mRNAs. We present here DPred, the first computational tool for predicting D on mRNAs in yeast from the primary RNA sequences. Built on a local self-attention layer and a convolutional neural network (CNN) layer, the proposed deep learning model outperformed classic machine learning approaches (random forest, support vector machines, etc.) and achieved reasonable accuracy and reliability with areas under the curve of 0.9166 and 0.9027 in jackknife cross-validation and on an independent testing dataset, respectively. Importantly, we showed that distinct sequence signatures are associated with the D sites on mRNAs and tRNAs, implying potentially different formation mechanisms and putative divergent functionality of this modification on the two types of RNA. DPred is available as a user-friendly Web server.

End-to-End Sentence-Level Multi-View Lipreading Architecture with Spatial Attention Module Integrated Multiple CNNs and Cascaded Local Self-Attention-CTC

Concomitant with the recent advances in deep learning, automatic speech recognition and visual speech recognition (VSR) have received considerable attention. However, although VSR systems must identify speech from both frontal and profile faces in real-world scenarios, most VSR studies have focused solely on frontal face pictures. To address this issue, we propose an end-to-end sentence-level multi-view VSR architecture for faces captured from four different perspectives (frontal, 30°, 45°, and 60°). The encoder uses multiple convolutional neural networks with a spatial attention module to detect minor changes in the mouth patterns of similarly pronounced words, and the decoder uses cascaded local self-attention connectionist temporal classification to collect the details of local contextual information in the immediate vicinity, which results in a substantial performance boost and speedy convergence. To compare the performance of the proposed model for experiments on the OuluVS2 dataset, the dataset was divided into four different perspectives, and the obtained performance improvement was 3.31% (0°), 4.79% (30°), 5.51% (45°), 6.18% (60°), and 4.95% (mean), respectively, compared with the existing state-of-the-art performance, and the average performance improved by 9.1% compared with the baseline. Thus, the suggested design enhances the performance of multi-view VSR and boosts its usefulness in real-world applications.