Search

As one of the most important post-translational modifications (PTMs), phosphorylation refers to the binding of a phosphate group with amino acid residues like Ser (S), Thr (T) and Tyr (Y) thus resulting in diverse functions at the molecular level. Abnormal phosphorylation has been proved to be closely related with human diseases. To our knowledge, no research has been reported describing specific disease-associated phosphorylation sites prediction which is of great significance for comprehensive understanding of disease mechanism. In this work, focusing on three types of leukemia, we aim to develop a reliable leukemia-related phosphorylation site prediction models by combing deep convolutional neural network (CNN) with transfer-learning. CNN could automatically discover complex representations of phosphorylation patterns from the raw sequences, and hence it provides a powerful tool for improvement of leukemia-related phosphorylation site prediction. With the largest dataset of myelogenous leukemia, the optimal models for S/T/Y phosphorylation sites give the AUC values of 0.8784, 0.8328 and 0.7716 respectively. When transferred learning on the small size datasets, the models for T-cell and lymphoid leukemia also give the promising performance by common sharing the optimal parameters. Compared with other five machine-learning methods, our CNN models reveal the superior performance. Finally, the leukemia-related pathogenesis analysis and distribution analysis on phosphorylated proteins along with K-means clustering analysis and position-specific conversation profiles on the phosphorylation site all indicate the strong practical feasibility of our easy-to-use CNN models.

Deep learning has proven to be a powerful method with applications in various fields including image, language, and biomedical data. Thanks to the libraries and toolkits such as TensorFlow, PyTorch, and Keras, researchers can use different deep learning architectures and data sets for rapid modeling. However, the available implementations of neural networks using these toolkits are usually designed for a specific research and are difficult to transfer to other work. Here, we present autoBioSeqpy, a tool that uses deep learning for biological sequence classification. The advantage of this tool is its simplicity. Users only need to prepare the input data set and then use a command line interface. Then, autoBioSeqpy automatically executes a series of customizable steps including text reading, parameter initialization, sequence encoding, model loading, training, and evaluation. In addition, the tool provides various ready-to-apply and adapt model templates to improve the usability of these networks. We introduce the application of autoBioSeqpy on three biological sequence problems: the prediction of type III secreted proteins, protein subcellular localization, and CRISPR/Cas9 sgRNA activity. autoBioSeqpy is freely available with examples at https://github.com/jingry/autoBioSeqpy.

Background: The aim of gene expression-based clinical modelling in tumorigenesis is not only to accurately predict the clinical endpoints, but also to reveal the genome characteristics for downstream analysis for the purpose of understanding the mechanisms of cancers. Most of the conventional machine learning methods involved a gene filtering step, in which tens of thousands of genes were firstly filtered based on the gene expression levels by a statistical method with an arbitrary cutoff. Although gene filtering procedure helps to reduce the feature dimension and avoid overfitting, there is a risk that some pathogenic genes important to the disease will be ignored., Results: In this study, we proposed a novel deep learning approach by combining a convolutional neural network with stationary wavelet transform (SWT-CNN) for stratifying cancer patients and predicting their clinical outcomes without gene filtering based on tumor genomic profiles. The proposed SWT-CNN overperformed the state-of-art algorithms, including support vector machine (SVM) and logistic regression (LR), and produced comparable prediction performance to random forest (RF). Furthermore, for all the cancer types, we firstly proposed a method to weight the genes with the scores, which took advantage of the representative features in the hidden layer of convolutional neural network, and then selected the prognostic genes for the Cox proportional-hazards regression. The results showed that risk stratifications can be effectively improved by using the identified prognostic genes as feature, indicating that the representative features generated by SWT-CNN can well correlate the genes with prognostic risk in cancers and be helpful for selecting the prognostic gene signatures., Conclusions: Our results indicated that gene expression-based SWT-CNN model can be an excellent tool for stratifying the prognostic risk for cancer patients. In addition, the representative features of SWT-CNN were validated to be useful for evaluating the importance of the genes in the risk stratification and can be further used to identify the prognostic gene signatures.

The accurate prediction of adverse drug reactions (ADRs) is essential for comprehensive drug safety evaluation. Pre-trained deep chemical language models have emerged as powerful tools capable of automatically learning molecular structural features from large-scale datasets, showing promising capabilities for the downstream prediction of molecular properties. However, the performance of pre-trained chemical language models in predicting ADRs, especially idiosyncratic ADRs induced by marketed drugs, remains largely unexplored. In this study, we propose MoLFormer-XL, a pre-trained model for encoding molecular features from canonical SMILES, in conjunction with a CNN-based model to predict drug-induced QT interval prolongation (DIQT), drug-induced teratogenicity (DIT), and drug-induced rhabdomyolysis (DIR). Our results demonstrate that the proposed model outperforms conventional models applied in previous studies for predicting DIQT, DIT, and DIR. Notably, an analysis of the learned linear attention maps highlights amines, alcohol, ethers, and aromatic halogen compounds as strongly associated with the three types of ADRs. These findings hold promise for enhancing drug discovery pipelines and reducing the drug attrition rate due to safety concerns. [ABSTRACT FROM AUTHOR]

Single-cell sequencing (scRNA-seq) technology provides higher resolution of cellular differences than bulk RNA sequencing and reveals the heterogeneity in biological research. The analysis of scRNA-seq datasets is premised on the subpopulation assignment. When an appropriate reference is not available, such as specific marker genes and single-cell reference atlas, unsupervised clustering approaches become the predominant option. However, the inherent sparsity and high-dimensionality of scRNA-seq datasets pose specific analytical challenges to traditional clustering methods. Therefore, a various deep learning-based methods have been proposed to address these challenges. As each method improves partially, a comprehensive method needs to be proposed. In this article, we propose a novel scRNA-seq data clustering method named AttentionAE-sc (Attention fusion AutoEncoder for single-cell). Two different scRNA-seq clustering strategies are combined through an attention mechanism, that include zero-inflated negative binomial (ZINB)-based methods dealing with the impact of dropout events and graph autoencoder (GAE)-based methods relying on information from neighbors to guide the dimension reduction. Based on an iterative fusion between denoising and topological embeddings, AttentionAE-sc can easily acquire clustering-friendly cell representations that similar cells are closer in the hidden embedding. Compared with several state-of-art baseline methods, AttentionAE-sc demonstrated excellent clustering performance on 16 real scRNA-seq datasets without the need to specify the number of groups. Additionally, AttentionAE-sc learned improved cell representations and exhibited enhanced stability and robustness. Furthermore, AttentionAE-sc achieved remarkable identification in a breast cancer single-cell atlas dataset and provided valuable insights into the heterogeneity among different cell subtypes. Author summary: Single-cell sequencing (scRNA-seq) has been widely used in numerous biological studies to reveal heterogeneity at the cellular level. Accurate cell type identification serves as the foundation for scRNA-seq data analysis, and unsupervised cluster analysis is commonly employed when an appropriate reference is not available. However, the inherent sparsity and high-dimensionality of scRNA-seq datasets pose specific analytical challenges to traditional clustering methods. To address this, we propose a novel scRNA-seq data clustering method named AttentionAE-sc (Attention fusion AutoEncoder for single-cell). By integrating denoising representation learning and cluster-friendly representation learning through an attention mechanism, AttentionAE-sc demonstrated outstanding performance in the evaluation phase. Firstly, when compared to benchmark methods on the real scRNA-seq datasets, AttentionAE-sc consistently achieved superior external and internal clustering evaluation metrics. Secondly, AttentionAE-sc exhibited robustness and stability across various experimental conditions. Lastly, AttentionAE-sc not only delivered excellent clustering results but also unveiled potential biological insights on a breast cancer single-cell atlas dataset. [ABSTRACT FROM AUTHOR]

Predicting RNA solvent accessibility using only primary sequence data can be regarded as sequence-based prediction work. Currently, the established studies for sequence-based RNA solvent accessibility prediction are limited due to the available number of datasets and black box prediction. To improve these issues, we first expanded the available RNA structures and then developed a sequence-based model using modified attention layers with different receptive fields to conform to the stem–loop structure of RNA chains. We measured the improvement with an extended dataset and further explored the model's interpretability by analysing the model structures, attention values and hyperparameters. Finally, we found that the developed model regarded the pieces of a sequence as templates during the training process. This work will be helpful for researchers who would like to build RNA attribute prediction models using deep learning in the future. [ABSTRACT FROM AUTHOR]

Cocrystal engineering have been widely applied in pharmaceutical, chemistry and material fields. However, how to effectively choose coformer has been a challenging task on experiments. Here we develop a graph neural network (GNN) based deep learning framework to quickly predict formation of the cocrystal. In order to capture main driving force to crystallization from 6819 positive and 1052 negative samples reported by experiments, a feasible GNN framework is explored to integrate important prior knowledge into end-to-end learning on the molecular graph. The model is strongly validated against seven competitive models and three challenging independent test sets involving pharmaceutical cocrystals, π–π cocrystals and energetic cocrystals, exhibiting superior performance with accuracy higher than 96%, confirming its robustness and generalization. Furthermore, one new energetic cocrystal predicted is successfully synthesized, showcasing high potential of the model in practice. All the data and source codes are available at https://github.com/Saoge123/ccgnet for aiding cocrystal community. Experimental determination of new cocrystals remains challenging due to the need of a systematic screening with a large range of coformers. Here the authors develop a flexible deep learning framework based on graph neural network demonstrated to quickly predict the formation of co-crystals. [ABSTRACT FROM AUTHOR]

[Display omitted] • Infrared data benefits herbal medicine detection. • Federated learning: privacy-preserving model training. • Deep learning: advanced neural networks for C.sinensis analysis. • Preprocessing: improves data quality for accurate analysis. The modeling and analysis of infrared spectrum data have gained significant advantages in the field of rapid detection of herbal medicine. However, extracting valuable insights and patterns from diverse chemical information relies on large-scale datasets and advanced algorithms. Unfortunately, due to privacy concerns, data sharing between different companies is often limited in practice. In such scenarios, federated learning emerges as a novel solution. Federated learning allows different regions or companies to jointly train models without sharing sensitive data, thereby protecting privacy. To address the quality control challenges related to Cordyceps sinensis(C.sinensis), this paper proposes a federated learning and deep learning method based on infrared spectroscopy data of C.sinensis, and then evaluates their performance on the C.sinensis authenticity detection and multi-classification tasks. In the classification tasks of 'part', 'phase', 'region', and 'True or False' (TorF), the mean MCC of the best model is 0.8805, 0.5865, 0.7004, and 0.9555 respectively. Experimental results show that the method proposed in this paper is an effective C.sinensis quality control method and it also provides solutions and ideas for privacy protection issues involving drug monitoring. [ABSTRACT FROM AUTHOR]

Transcription factors are proteins that could bind to specific DNA sequences so as to regulate gene expressions. Currently, identification of transcription factor binding sites locating in DNA sequences is very important for building regulatory model in biological systems and identifying pathogenic variations. Traditional machine-learning methods have been successfully used for biological prediction problems based on DNA or protein sequences, but they all need to manually extract numerical features, which is not only tedious, but also would ignore effective information of first-order sequences. In this paper, based on the principle of deep learning (DL), we constructed prediction model for transcription factor binding sites only from DNA original base sequences. Here, a DL method based on convolutional neural network (CNN) and long short-term memory (LSTM) were proposed to investigate four leukemia categories from the perspective of transcription factor binding sites using four large non-redundant datasets for acute, chronic, myeloid and lymphatic leukemia, respectively. Compared with three widely used machine-learning methods of artificial neural network (ANN), support vector machine (SVM) and random forest (RF), our DL method exhibits significant superiority in terms of prediction performance, since the prediction accuracy of three machine-learning models either based on sequence feature or k-mer feature extraction are all lower than that of DL model. The available DL models for four leukemia categories gives an average prediction accuracy of 75% based only on sequence segments with 101 bases, which indicates that the DL based method is promising with unique advantages over the traditional machine learning methods. But focusing on leukemia-related transcription factor binding site prediction, further improvements would be implemented such as optimizing base segment length and CNN architecture, in order to improve the current prediction accuracy. • A new deep learning method was firstly proposed for predicting transcription factor (TFs) binding sites only based on the original DNA sequences. • The method could better use the original sequence and location information without considering the optimization of effective features. • By compared with three traditional machine-learning methods, the CNN method gives the best performance either based on sequence feature or k-mer feature extraction. • The results indicate that deep learning has the great potential in solving prediction problems from biological sequences. [ABSTRACT FROM AUTHOR]

Post-translational modifications of histone play a key role in controlling gene expression, as they are involved in gene transcription regulation through controlling DNA accessibility or recruiting specific transcription factors. Developing computational methods to predict gene expression levels from histone modification data may facilitate understanding their function in gene regulation and contribute to develop 'epigenetic drugs' for disease treatment. Several studies have reported that transcription termination plays a crucial role in gene regulation as it is able to premature termination and determine the cellular fate of transcripts. However, previous work mainly focused on the histone characteristics flanking transcription start site (TSS) of genes, while ignoring the histone modification features around transcription termination site (TTS). Thus, we introduced a hybrid convolutional and bi-directional long short-term memory network with attention mechanism to predict gene expression from histone modification signals in both regions. Our model achieved higher results than the state-of-the-art model in both classification and regression tasks to infer gene expression level. The predictive results of model demonstrated that histone signals of TTS can provide additional information to improve the model performance. In addition, attention weights and transcription factor binding data indicated that histone modifications in TSS might regulate transcription by recruiting TFs, but the regulation mechanisms of it in TTS need to be further explored. • A new hybrid deep learning model was proposed for predicting gene expression levels and achieved better performance compared to baselines. • The model integrated histone features from both TSS and TTS region and validated the function role of histone modification of TTS in transcription regulation. • We explored the possible regulation mechanism of histone modifications in TSS and TTS. [ABSTRACT FROM AUTHOR]