Abstract:The precise identification of splice sites is essential for unraveling the structure and function of genes, constituting a pivotal step in the gene annotation process. In this study, we developed a novel deep learning model, DRANetSplicer, that integrates residual learning and attention mechanisms for enhanced accuracy in capturing the intricate features of splice sites. We constructed multiple datasets using the most recent versions of genomic data from three different organisms, Oryza sativa japonica, Arabidopsis thaliana and Homo sapiens. This approach allows us to train models with a richer set of high-quality data. DRANetSplicer outperformed benchmark methods on donor and acceptor splice site datasets, achieving an average accuracy of (96.57%, 95.82%) across the three organisms. Comparative analyses with benchmark methods, including SpliceFinder, Splice2Deep, Deep Splicer, EnsembleSplice, and DNABERT, revealed DRANetSplicer's superior predictive performance, resulting in at least a (4.2%, 11.6%) relative reduction in average error rate. We utilized the DRANetSplicer model trained on O. sativa japonica data to predict splice sites in A. thaliana, achieving accuracies for donor and acceptor sites of (94.89%, 94.25%). These results indicate that DRANetSplicer possesses excellent cross-organism predictive capabilities, with its performance in cross-organism predictions even surpassing that of benchmark methods in non-cross-organism predictions. Cross-organism validation showcased DRANetSplicer's excellence in predicting splice sites across similar organisms, supporting its applicability in gene annotation for understudied organisms. We employed multiple methods to visualize the decision-making process of the model. The visualization results indicate that DRANetSplicer can learn and interpret well-known biological features, further validating its overall performance. Our study systematically examined and confirmed the predictive ability of DRANetSplicer from various levels and perspectives, indicating that its practical application in gene annotation is justified.

Predicting RNA splicing from DNA sequence using Pangolin

LncLSTA: A Versatile Predictor Unveiling Subcellular Localization of Lncrnas Through Long-Short Term Attention

Splam: a deep-learning-based splice site predictor that improves spliced alignments

Reference-informed prediction of alternative splicing and splicing-altering mutations from sequences

A Deep Learning-Based Sequence Analyzer Incorporating the Transcription Factor Binding Affinity to Dissect the Effects of Non-Coding Genetic Variants

SpliceRover: interpretable convolutional neural networks for improved splice site prediction

A New Method for Splice Site Prediction Based on the Sequence Patterns of Splicing Signals and Regulatory Elements

DRANetSplicer: A Splice Site Prediction Model Based on Deep Residual Attention Networks

Predicting splicing patterns from the transcription factor binding sites in the promoter with deep learning

SpliceTransformer predicts tissue-specific splicing linked to human diseases

Self-supervised learning on millions of primary RNA sequences from 72 vertebrates improves sequence-based RNA splicing prediction

Machine learning-optimized targeted detection of alternative splicing

Prediction of Primate Splice Site Using Inhomogeneous Markov Chain and Neural Network.

DeepSplice: a deep learning approach for accurate prediction of alternative splicing events in the human genome

CADD-Splice—improving genome-wide variant effect prediction using deep learning-derived splice scores

PLEKv2: predicting lncRNAs and mRNAs based on intrinsic sequence features and the coding-net model

DeepSS: Exploring Splice Site Motif Through Convolutional Neural Network Directly from DNA Sequence

Improved Rat Genome Gene Prediction by Integration of ESTs with RNA-Seq Information

An effective deep learning-based approach for splice site identification in gene expression

Advances in alternative splicing identification: deep learning and pantranscriptome