Structure-aware deep learning enhances m6A prediction and reveals cell type-associated RNA structural signatures

PLM-ArgMe: Protein language model for arginine methylation prediction for different species.

PLM-ArgMe is presented that is based on a symmetry-sensitive Transformer framework using context-aware ESM-2 residue embeddings, which is mapped through a novel Bio-Symmetric Mirrored Sinusoidal Encoding strategy to address the biological symmetry hypothesis of arginine methylation.

A scalable deep learning framework for RNA 5-methyluridine site prediction using hybrid features and high-performance parallel computing

M2-PRNet: multi-scale and multi-modal learning for protein–RNA binding affinity prediction

The results demonstrate the effectiveness of integrating multi-scale and multi-modal representations with cross-scale alignment for protein–RNA affinity prediction, and suggest that M2-PRNet can highlight relevant RNA-binding regions and support preliminary discrimination between strong and weak binders when plausible complex structures are available.

Deep3MVPF: Multiview Deep Framework for the Prediction of Stability and m6A in mRNA 3'UTR.

Deep3MVPF, a multiview deep learning framework for 3'UTR stability prediction and m6A site identification, integrates a multiscale convolutional neural network, a k-mer de Bruijn graph neural network, and a secondary-structure graph neural network to jointly model sequence, topological, and structural representations.

BLOSSOM-Kcr: A structure-informed deep learning framework for lysine crotonylation site prediction.

Lysine crotonylation (Kcr) is an important post-translational modification (PTM) involved in diverse biological processes, including chromatin regulation, protein function modulation, and cellular signaling. Although mass spectrometry-based proteomics has substantially expanded the identification of Kcr sites, experimental screening remains labor-intensive, costly, and difficult to apply at proteome scale. Computational methods provide an efficient strategy for prioritizing candidate Kcr sites. However, most existing predictors mainly rely on sequence-derived representations and insufficiently exploit protein structural context. In this study, we propose BLOSSOM-Kcr, a structure-informed deep learning framework for Kcr site prediction. BLOSSOM-Kcr integrates BLOSUM62-based sequence substitution features with residue-level structural descriptors, including secondary structure, solvent accessibility, backbone geometry, and spatial neighborhood information. The fused residue-level representation is further processed by residual convolutional blocks, channel attention, bidirectional long short-term memory (BiLSTM) layers, and attention pooling to capture local motif patterns, informative feature dimensions, and contextual dependencies surrounding candidate lysine residues. Fivefold cross-validation was performed for model optimization and comparative analysis, while an independent test set was used for final evaluation against existing Kcr site predictors. On the independent test set, BLOSSOM-Kcr achieved an AUC of 0.9023, an MCC of 0.6479, and an F1-score of 0.8336, outperforming representative Kcr site predictors. These results suggest that BLOSSOM-Kcr provides an effective structure-aware framework for Kcr site prediction.

DNCLA: A Deep Learning Model for TFBS Identification Based on Structural and Conformational Properties of Nucleotides and Dinucleotides

Identifying transcription factor binding sites (TFBSs) is fundamental to understanding complex gene regulatory mechanisms and the functions of non-coding regions. Although existing methods have achieved substantial strides, capturing both local structural features and long-range spatial dependencies within DNA sequences remains a major challenge for improving prediction accuracy. In this study, we propose DNCLA, a deep learning model that synergizes multisize convolutional fusion, Bidirectional Long ShortTerm Memory (Bi-LSTM) networks, and a multi-head self-attention mechanism. At the feature extraction level, DNCLA breaks through the limitations of traditional single-sequence encoding by fusing Nucleotide Chemical Properties (NCP) with Dinucleotide Physicochemical Properties (DPCP). NCP provides a refined characterization of chemical differences between bases based on ring structures, hydrogen bond sites, and functional group properties, while DPCP introduces parameters such as local structural stability and geometric flexibility of the DNA. Subsequently, the model extracts spatial evolution from these high-dimensional features through a multi-size convolutional module; captures long-range spatial dependencies using Bi-LSTM layers; and employs a multi-head self-attention mechanism to achieve adaptive weight distribution of global features, thereby enhancing the perception of key regulatory motifs. Results from training and testing the proposed model on 165 ChIPseq datasets demonstrate that DNCLA possesses robust generalization capabilities and high predictive performance in TFBSs identification. This suggests that the incorporation of physicochemical features better elucidates the essence of interactions between transcription factors and DNA.