Encoder transformer models compress information from all tokens in a sequence into a single [CLS] token to represent global context.This approach risks diluting fine-grained or hierarchical features, leading to information loss in downstream tasks where local patterns are important.To remedy this, we propose a lightweight architectural enhancement: an inception-style 1-D convolution module that sits on top of the transformer layer and augments token representations with multi-scale local features.This enriched feature space is then processed by a self-attention layer that dynamically weights tokens based on their task relevance.Experiments on five diverse tasks show that our framework consistently improves general-purpose, domain-specific, and multilingual models, outperforming baselines by 1% to 14% while maintaining efficiency.Ablation studies show that multi-scale convolution performs better than any single kernel and that the self-attention layer is critical for performance.

The subcellular localization of messenger RNAs (mRNAs) plays a crucial role in gene regulation, ensuring precise spatial and temporal control of protein synthesis. Traditional computational approaches for mRNA localization have primarily relied on single-label classification models, which fail to capture the inherent multi-label nature of mRNA localization. Recent advancements have introduced deep learning-based multi-label prediction frameworks; however, existing methods often lack an effective way to model the relationships between multiple localizations. In this paper, we propose Localization with Supervised Contrastive Learning (LOCAS), a novel approach for multi-label mRNA subcellular localization prediction. LOCAS integrates an RNA language model (RiNALMo) to generate high-quality sequence embeddings and employs supervised contrastive learning (SCL) to refine the embedding space, ensuring biologically meaningful clustering of RNA sequences. To handle overlapping labels, we introduce an overlap-threshold-based similarity measure during contrastive training. Finally, we leverage an ML-Decoder, which utilizes a cross-attention mechanism to enhance multi-label classification performance. We evaluate LOCAS on two benchmark datasets, RNALocate and RNALocate V2.0, demonstrating state-of-the-art performance across all evaluation metrics. Extensive ablation studies validate the effectiveness of our approach, highlighting the contributions of contrastive learning and ML-decoder in improving multi-label classification. Our results suggest that integrating RNA sequence representation learning with SCL offers a powerful and scalable solution for mRNA localization prediction.

RNA torsion and pseudo-torsion angles are critical in determining the three-dimensional conformation of RNA molecules, which in turn governs their biological functions. However, current methods are limited by RNA's structural complexity as well as flexibility, with experimental techniques being costly and computational approaches struggling to capture the intricate sequence dependencies needed for accurate predictions. To address these challenges, we introduce DeepRNA-Twist, a novel deep learning framework designed to predict RNA torsion and pseudo-torsion angles directly from sequence. DeepRNA-Twist utilizes RNA language model embeddings, which provides rich, context-aware feature representations of RNA sequences. Additionally, it introduces 2A3IDC module (Attention Augmented Inception Inside Inception with Dilated CNN), combining inception networks with dilated convolutions and multi-head attention mechanism. The dilated convolutions capture long-range dependencies in the sequence without requiring a large number of parameters, while the multi-head attention mechanism enhances the model's ability to focus on both local and global structural features simultaneously. DeepRNA-Twist was rigorously evaluated on benchmark datasets, including RNA-Puzzles, CASP-RNA, and SPOT-RNA-1D, and demonstrated significant improvements over existing methods, achieving state-of-the-art accuracy. Source code is available at https://github.com/abrarrahmanabir/DeepRNA-Twist.

BACKGROUND: DNA-binding proteins (DNA-BPs) are the proteins that bind and interact with DNA. DNA-BPs regulate and affect numerous biological processes, such as, transcription and DNA replication, repair, and organization of the chromosomal DNA. Very few proteins, however, are DNA-binding in nature. Therefore, it is necessary to develop an efficient predictor for identifying DNA-BPs. RESULT: In this work, we have proposed new benchmark datasets for the DNA-binding protein prediction problem. We discovered several quality concerns with the widely used benchmark datasets, PDB1075 (for training) and PDB186 (for independent testing), which necessitated the preparation of new benchmark datasets. Our proposed datasets UNIPROT1424 and UNIPROT356 can be used for model training and independent testing respectively. We have retrained selected state-of-the-art DNA-BP predictors in the new dataset and reported their performance results. We also trained a novel predictor using the new benchmark dataset. We extracted features from various feature categories, then used a Random Forest classifier and Recursive Feature Elimination with Cross-validation (RFECV) to select the optimal set of 452 features. We then proposed a stacking ensemble architecture as our final prediction model. Named Stacking Ensemble Model for DNA-binding Protein Prediction, or StackDPP in short, our model achieved 0.92, 0.92 and 0.93 accuracy in 10-fold cross-validation, jackknife and independent testing respectively. CONCLUSION: StackDPP has performed very well in cross-validation testing and has outperformed all the state-of-the-art prediction models in independent testing. Its performance scores in cross-validation testing generalized very well in the independent test set. The source code of the model is publicly available at https://github.com/HasibAhmed1624/StackDPP . Therefore, we expect this generalized model can be adopted by researchers and practitioners to identify novel DNA-binding proteins.

Lysine succinylation is a kind of post-translational modification (PTM) that plays a crucial role in regulating the cellular processes. Aberrant succinylation may cause inflammation, cancers, metabolism diseases and nervous system diseases. The experimental methods to detect succinylation sites are time-consuming and costly. This thus calls for computational models with high efficacy, and attention has been given in the literature to develop such models, albeit with only moderate success in the context of different evaluation metrics. One crucial aspect in this context is the biochemical and physicochemical properties of amino acids, which appear to be useful as features for such computational predictors. However, some of the existing computational models did not use the biochemical and physicochemical properties of amino acids. In contrast, some others used them without considering the inter-dependency among the properties. The combinations of biochemical and physicochemical properties derived through our optimization process achieve better results than the results achieved by combining all the properties. We propose three deep learning architectures: CNN+Bi-LSTM (CBL), Bi-LSTM+CNN (BLC) and their combination (CBL_BLC). We find that CBL_BLC outperforms the other two. Ensembling of different models successfully improves the results. Notably, tuning the threshold of the ensemble classifiers further improves the results. Upon comparing our work with other existing works on two datasets, we successfully achieve better sensitivity and specificity by varying the threshold value.