RNA pseudouridine modification is particularly important in a variety of cellular biological and physiological processes. It plays a significant role in understanding RNA functions, RNA structure stabilization, translation processes, etc. To understand its functional mechanisms, it is necessary to accurately identify pseudouridine sites in RNA sequences. Although some computational methods have been proposed for the identification of pseudouridine sites, it is still a challenge to improve the identification accuracy and generalization ability. To address this challenge, a novel feature fusion predictor, named PsoEL-PseU, is proposed for the prediction of pseudouridine sites. Firstly, this study systematically and comprehensively explored different types of feature descriptors and determined six feature descriptors with various properties. To improve the feature representation ability, a binary particle swarm optimizer was used to capture the optimal feature subset for six feature descriptors. Secondly, six individual predictors were trained by using the six optimal feature subsets. Finally, to fuse the effects of all six features, six individual predictors were fused into an ensemble predictor by a parallel fusion strategy. Ten-fold cross-validation on three benchmark datasets indicated that the PsoEL-PseU predictor significantly outperformed the current state-of-the-art predictors. Additionally, the new predictor achieved better accuracy in the independent dataset evaluation—accuracy which is significantly higher than that of its existing counterparts—and the user-friendly webserver developed by the PsoEL-PseU predictor has been made freely accessible.

The subcellular localization of messenger RNA (mRNA) is essential for the regulation of gene expression and plays a pivotal role in targeted drug development. Although several computational models have been developed to predict mRNA localization, these approaches still face challenges in sequence representation and exhibit limited performance in handling multi-localization tasks. In this paper, we propose mRSubLoc, a novel multi-label deep learning framework for predicting mRNA subcellular localization. The model integrates the RNA large language model RNAErnie with one-hot encoding and Word2Vec embeddings to construct a comprehensive representation of mRNA sequences. A text convolutional neural network (TextCNN) is employed to capture local feature patterns, while a bidirectional long short-term memory network (BiLSTM) is used to capture long-range dependencies. These features are fused using a multi-head self-attention mechanism to effectively capture localization-specific characteristics. Finally, a multi-layer perceptron (MLP) explores complex dependencies among multiple localization sites, facilitating accurate mRNA subcellular localization prediction. Experimental results on a testing set demonstrate that mRSubLoc significantly outperforms state-of-the-art methods across multiple metrics, including Aiming (0.7858), Coverage (0.6212), Accuracy (0.6161), Absolute-True (0.3070), and Absolute-False (0.1319). This study proposes a novel approach for predicting mRNA subcellular localization and provides new perspectives for advancing disease diagnosis and drug discovery in biomedical research.