Search

Enhancers are crucial cis-regulatory elements that control gene expression in a cell-type-specific manner. Despite extensive genetic and computational studies, accurately predicting enhancer activity in different cell types remains a challenge, and the grammar of enhancers is still poorly understood. Here, we present HEAP (high-resolution enhancer activity prediction), an explainable deep learning framework for predicting enhancers and exploring enhancer grammar. The framework includes three modules that use grammar-based reasoning for enhancer prediction. The algorithm can incorporate DNA sequences and epigenetic modifications to obtain better accuracy. We use a novel two-step multi-task learning method, task adaptive parameter sharing (TAPS), to efficiently predict enhancers in different cell types. We first train a shared model with all cell-type datasets. Then we adapt to specific tasks by adding several task-specific subset layers. Experiments demonstrate that HEAP outperforms published methods and showcases the effectiveness of the TAPS, especially for those with limited training samples. Notably, the explainable framework HEAP utilizes post-hoc interpretation to provide insights into the prediction mechanisms from three perspectives: data, model architecture and algorithm, leading to a better understanding of model decisions and enhancer grammar. To the best of our knowledge, HEAP will be a valuable tool for insight into the complex mechanisms of enhancer activity., (© The Author(s) 2023. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com.)

Recent advances in single-cell sequencing assays for the transposase-accessibility chromatin (scATAC-seq) technique have provided cell-specific chromatin accessibility landscapes of cis-regulatory elements, providing deeper insights into cellular states and dynamics. However, few research efforts have been dedicated to modeling the relationship between regulatory grammars and single-cell chromatin accessibility and incorporating different analysis scenarios of scATAC-seq data into the general framework. To this end, we propose a unified deep learning framework based on the ProdDep Transformer Encoder, dubbed PROTRAIT, for scATAC-seq data analysis. Specifically motivated by the deep language model, PROTRAIT leverages the ProdDep Transformer Encoder to capture the syntax of transcription factor (TF)-DNA binding motifs from scATAC-seq peaks for predicting single-cell chromatin accessibility and learning single-cell embedding. Based on cell embedding, PROTRAIT annotates cell types using the Louvain algorithm. Furthermore, according to the identified likely noises of raw scATAC-seq data, PROTRAIT denoises these values based on predated chromatin accessibility. In addition, PROTRAIT employs differential accessibility analysis to infer TF activity at single-cell and single-nucleotide resolution. Extensive experiments based on the Buenrostro2018 dataset validate the effeteness of PROTRAIT for chromatin accessibility prediction, cell type annotation, and scATAC-seq data denoising, therein outperforming current approaches in terms of different evaluation metrics. Besides, we confirm the consistency between the inferred TF activity and the literature review. We also demonstrate the scalability of PROTRAIT to analyze datasets containing over one million cells.

Currently, the Thinprep cytologic test (TCT) is the most popular cervical cancer cytology test technique. It can detect precancerous conditions and microbial infections. However, this technique entirely relies on manual operation and doctors' naked eye observation, resulting in a heavy workload and low accuracy rate. Recently, automatic pathological diagnosis has been developed to solve this problem. Cervical cell classification is a key technology in the intelligent cervical cancer diagnosis system. Training a deep neural network-based classification model requires a large amount of data. However, cervical cell labeling requires specialized physicians and the cost of labeling is high, resulting in a lack of sufficient labeling data in this field. To address this problem, we propose a method to ensure high accuracy in cervical cell classification with a small amount of labeled data by introducing manual features and a voting mechanism to achieve data expansion in semi-supervised learning. The method consists of three main steps, using a clarity function to filter out high-quality cervical cell images, annotating a small amount of them, and balancing the training data using a voting mechanism. With a small amount of labeled data, the accuracy of the proposed method in this paper can reach to 91.94%., Competing Interests: The authors declare that there is no conflict of interest regarding the publication of this paper., (Copyright © 2022 Siqi Zhao et al.)

Transcription factors (TFs) are essential proteins in regulating the spatiotemporal expression of genes. It is crucial to infer the potential transcription factor binding sites (TFBSs) with high resolution to promote biology and realize precision medicine. Recently, deep learning-based models have shown exemplary performance in the prediction of TFBSs at the base-pair level. However, the previous models fail to integrate nucleotide position information and semantic information without noisy responses. Thus, there is still room for improvement. Moreover, both the inner mechanism and prediction results of these models are challenging to interpret. To this end, the Deep Attentive Encoder-Decoder Neural Network (D-AEDNet) is developed to identify the location of TFs-DNA binding sites in DNA sequences. In particular, our model adopts Skip Architecture to leverage the nucleotide position information in the encoder and removes noisy responses in the information fusion process by Attention Gate. Simultaneously, the Transcription Factor Motif Discovery based on Sliding Window (TF-MoDSW), an approach to discover TFs-DNA binding motifs by utilizing the output of neural networks, is proposed to understand the biological meaning of the predicted result. On ChIP-exo datasets, experimental results show that D-AEDNet has better performance than competing methods. Besides, we authenticate that Attention Gate can improve the interpretability of our model by ways of visualization analysis. Furthermore, we confirm that ability of D-AEDNet to learn TFs-DNA binding motifs outperform the state-of-the-art methods and availability of TF-MoDSW to discover biological sequence motifs in TFs-DNA interaction by conducting experiment on ChIP-seq datasets., (© The Author(s) 2021. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com.)

In air traffic control (ATC), speech communication with radio transmission is the primary way to exchange information between the controller and the pilot. As a result, the integration of automatic speech recognition (ASR) systems holds immense potential for reducing controllers' workload and plays a crucial role in various ATC scenarios, which is particularly significant for ATC research. This article provides a comprehensive review of ASR technology's applications in the ATC communication system. Firstly, it offers a comprehensive overview of current research, including ATC corpora, ASR models, evaluation measures and application scenarios. A more comprehensive and accurate evaluation methodology tailored for ATC is proposed, considering advancements in communication sensing systems and deep learning techniques. This methodology helps researchers in enhancing ASR systems and improving the overall performance of ATC systems. Finally, future research recommendations are identified based on the primary challenges and issues. The authors sincerely hope this work will serve as a clear technical roadmap for ASR endeavors within the ATC domain and make a valuable contribution to the research community. [ABSTRACT FROM AUTHOR]

In recent years, the civil aviation industry has actively promoted the automation and intelligence of control processes with the increasing use of various artificial intelligence technologies. Air–ground communication, as the primary means of interaction between controllers and pilots, typically involves one or more intents. Recognizing multiple intents within air–ground communication texts is a critical step in automating and advancing the control process intelligently. Therefore, this study proposes a hybrid detection method for multi-intent recognition in air–ground communication text. This method improves recognition accuracy by using different models for single-intent texts and multi-intent texts. First, the air–ground communication text is divided into two categories using multi-intent detection technology: single-intent text and multi-intent text. Next, for single-intent text, the Enhanced Representation through Knowledge Integration (ERNIE) 3.0 model is used for recognition; while the A Lite Bidirectional Encoder Representations from Transformers (ALBERT)_Sequence-to-Sequence_Attention (ASA) model is proposed for identifying multi-intent texts. Finally, combining the recognition results from the two models yields the final result. Experimental results demonstrate that using the ASA model for multi-intent text recognition achieved an accuracy rate of 97.84%, which is 0.34% higher than the baseline ALBERT model and 0.15% to 0.87% higher than other improved models based on ALBERT and ERNIE 3.0. The single-intent recognition model achieved an accuracy of 96.23% when recognizing single-intent texts, which is at least 2.18% higher than the multi-intent recognition model. The results indicate that employing different models for various types of texts can substantially enhance recognition accuracy. [ABSTRACT FROM AUTHOR]

Against the backdrop of global warming and increased rainfall, the hazards and potential risks of landslides are increasing. The rapid generation of a landslide inventory is of great significance for landslide disaster prevention and reduction. Deep learning has been widely applied in landslide identification due to its advantages in terms of its deeper model structure, high efficiency, and high accuracy. This article first provides an overview of deep learning technology and its basic principles, as well as the current status of landslide remote sensing databases. Then, classic landslide deep learning recognition models such as AlexNet, ResNet, YOLO, Mask R-CNN, U-Net, Transformer, EfficientNet, DeeplabV3+ and PSPNet were introduced, and the advantages and limitations of each model were extensively analyzed. Finally, the current constraints of deep learning in landslide identification were summarized, and the development direction of deep learning in landslide identification was analyzed. The purpose of this article is to promote the in-depth development of landslide identification research in order to provide academic references for the prevention and mitigation of landslide disasters and post-disaster rescue work. The research results indicate that deep learning methods have the characteristics of high efficiency and accuracy in automatic landslide recognition, and more attention should be paid to the development of emerging deep learning models in landslide recognition in the future. [ABSTRACT FROM AUTHOR]

Retraction: [Xiaojian Wang, Zixuan Wang, Xiaoye Sun, Multi‐mode urban rail transit and spatial coordinated development based on deep learning system, IET Circuits, Devices & Systems 2023 (https://doi.org/10.1049/cds2.12144)]. The above article from IET Circuits, Devices & Systems, published online on 10 January 2023 in Wiley Online Library (wileyonlinelibrary.com), has been retracted by agreement between the Editor‐in‐Chief, Harry E. Ruda, the Institution of Engineering and Technology (the IET) and John Wiley and Sons Ltd. This article was published as part of a Guest Edited special issue. Following an investigation, the IET and the journal have determined that the article was not reviewed in line with the journal's peer review standards and there is evidence that the peer revie process of the special issue underwent systematic manipulation. Accordingly, we cannot vouch for the integrity or reliability of the content. As such we have taken the decision to retract the article. The authors have been informed of the decision to retract. [ABSTRACT FROM AUTHOR]

Chromatin features can reveal tissue-specific TF-DNA binding, which leads to a better understanding of many critical physiological processes. Accurately identifying TF-DNA bindings and constructing their relationships with chromatin features is a long-standing goal in the bioinformatic field. However, this has remained elusive due to the complex binding mechanisms and heterogeneity among inputs. Here, we have developed the GHTNet (General Hybrid Transformer Network), a transformer-based model to predict TF-DNA binding specificity. The GHTNet decodes the relationship between tissue-specific TF-DNA binding and chromatin features via a specific input scheme of alternative inputs and reveals important gene regions and tissue-specific motifs. Our experiments show that the GHTNet has excellent performance, achieving about a 5% absolute improvement over existing methods. The TF-DNA binding mechanism analysis shows that the importance of TF-DNA binding features varies across tissues. The best predictor is based on the DNA sequence, followed by epigenomics and shape. In addition, cross-species studies address the limited data, thus providing new ideas in this case. Moreover, the GHTNet is applied to interpret the relationship among TFs, chromatin features, and diseases associated with AD46 tissue. This paper demonstrates that the GHTNet is an accurate and robust framework for deciphering tissue-specific TF-DNA binding and interpreting non-coding regions. [ABSTRACT FROM AUTHOR]

Among the existing bearing fault diagnosis algorithms, the diagnosis time of algorithms with high accuracy is relatively long, and the accuracy of some lightweight methods is low compared to other methods. This work proposes a bi-level binary coded fully connected classifier based on residual network 50 with bottom and deep level features (Bi-BUR) for bearing fault diagnosis, enabling both high accuracy and high-speed characteristics. The Bi-BUR has two levels of classifiers. The first level begins with a signal-to-picture method based on integer binary coding to convert the bearing fault signals into binary pictures; then, the first classification of the converted pictures is performed by the underlying feature extraction-residual network 50, which can extract both the underlying and the advanced features. The second level performs modal decomposition of the signals that are not classified successfully in the first level, performs the second classification with a fully connected classifier inserted into the underlying feature extraction module, and exports the final classification results after weighted summation of the results of each type. The Bi-BUR is simulated on the Case Western Reserve University motor, the self-priming centrifugal pump, and the axial piston hydraulic pump bearing failure datasets. All three experiments achieved more than 95% accuracy in the first level of categorization and 100% diagnosis after the second level of categorization. And the Bi-BUR is faster than other high-accuracy methods. • A bearing fault diagnosis method with fusion of speed and accuracy is considered. • A two-layer bearing fault diagnosis method is proposed. • A signal-to-picture method based on integer binary coding is proposed. • A method to extract both bottom and deep level features is proposed. • The method fills the gap of bearing fault diagnosis with high speed and accuracy. [ABSTRACT FROM AUTHOR]

Traffic Classification (TC) has been applied to a wide range of applications, from security monitoring to quality of service (QoS) provisioning in network Internet Service Providers (ISPs). In recent years, many researchers have applied Machine Learning (ML) or Deep Learning (DL) to TC, namely AI-TC. However, AI-TC methods face significant challenges, including high data dependency, exhaustively costly traffic labeling, and network subscribers' privacy. This paper proposes a TC framework for smart home networks using Federated Learning (FL) that protects traffic data privacy by performing local training and inference of TC models. Firstly, we design a DPI-based traffic labeling method on edge home gateways as FL nodes, which enables these nodes to have data labeling capability while protecting data privacy. Then, a semi-supervised TC model based on an autoencoder (AE) is proposed to reduce the dependence of the model on labeled traffic samples. Finally, an XAI-based method is utilized to interpret the model to ensure its explainability. We validate the proposed method on public and real datasets using benchmarking methods. The experimental results show that the method can achieve high performance using a small number of samples while protecting data privacy and improving the model's credibility. Experimental code can be found in the following url: https://github.com/PrinceXuan12138/HGW-TC-Experimental-code. [ABSTRACT FROM AUTHOR]

Currently, the Thinprep cytologic test (TCT) is the most popular cervical cancer cytology test technique. It can detect precancerous conditions and microbial infections. However, this technique entirely relies on manual operation and doctors' naked eye observation, resulting in a heavy workload and low accuracy rate. Recently, automatic pathological diagnosis has been developed to solve this problem. Cervical cell classification is a key technology in the intelligent cervical cancer diagnosis system. Training a deep neural network-based classification model requires a large amount of data. However, cervical cell labeling requires specialized physicians and the cost of labeling is high, resulting in a lack of sufficient labeling data in this field. To address this problem, we propose a method to ensure high accuracy in cervical cell classification with a small amount of labeled data by introducing manual features and a voting mechanism to achieve data expansion in semi-supervised learning. The method consists of three main steps, using a clarity function to filter out high-quality cervical cell images, annotating a small amount of them, and balancing the training data using a voting mechanism. With a small amount of labeled data, the accuracy of the proposed method in this paper can reach to 91.94%. [ABSTRACT FROM AUTHOR]

Summary: The microscopic image is important data for recording the microstructure information of materials. Researchers usually use image‐processing algorithms to extract material features from that and then characterise the material microstructure. However, the microscopic images obtained by a microscope often have random damaged regions, which will cause the loss of information and thus inevitably influence the accuracy of microstructural characterisation, even lead to a wrong result. To handle this problem, we provide a deep learning‐based fully automatic method for detecting and inpainting damaged regions in material microscopic images, which can automatically inpaint damaged regions with different positions and shapes, as well as we also use a data augmentation method to improve the performance of inpainting model. We evaluate our method on Al–La alloy microscopic images, which indicates that our method can achieve promising performance on inpainted and material microstructure characterisation results compared to other image inpainting software for both accuracy and time consumption. Lay Description: A basic goal of materials data analysis is to extract useful information from materials datasets that can in turn be used to establish connections along the composition–processing–structure–properties chain. The microscopic images obtained by a microscope is the key carrier of material microstructural information. Researchers usually use image analysis algorithms to extract regions of interest or useful features from microscopic images, aiming to analyse material microstructure, organ tissues or device quality etc. Therefore, the integrity and clarity of the microscopic image are the most important attributes for image feature extraction. Scientists and engineers have been trying to develop various technologies to obtain perfect microscopic images. However, in practice, some extrinsic defects are often introduced during the preparation and/or shooting processes, and the elimination of these defects often requires mass efforts and cost, or even is impossible at present. Take the microstructure image of metallic material for example, samples prepared to microstructure characterisation often need to go through several steps such as cutting, grinding with sandpaper, polishing, etching, and cleaning. During the grinding and polishing process, defects such as scratches could be introduced. During the etching and cleaning process, some defects such as rust caused by substandard etching, stains etc. may arise and be persisted. These defects can be treated as damaged regions with nonfixed positions, different sizes, and random shapes, resulting in the loss of information, which seriously affects subsequent visual observation and microstructural feature extraction. To handle this problem, we provide a deep learning‐based fully automatic method for detecting and inpainting damaged regions in material microscopic images, which can automatically inpaint damaged regions with different positions and shapes, as well as we also use a data augmentation method to improve the performance of inpainting model. We evaluate our method on Al–La alloy microscopic images, which indicates that our method can achieve promising performance on inpainted and material microstructure characterisation results compared to other image inpainting software for both accuracy and time consumption. [ABSTRACT FROM AUTHOR]

Land cover validation plays an important role in the process of generating and distributing land cover thematic maps, which is usually implemented by high cost of sample interpretation with remotely sensed images or field survey. With an increasing availability of geo-tagged landscape photos, the automatic photo recognition methodologies, e.g., deep learning, can be effectively utilised for land cover applications. However, they have hardly been utilised in validation processes, as challenges remain in sample selection and classification for highly heterogeneous photos. This study proposed an approach to employ geo-tagged photos for land cover validation by using the deep learning technology. The approach first identified photos automatically based on the VGG-16 network. Then, samples for validation were selected and further classified by considering photos distribution and classification probabilities. The implementations were conducted for the validation of the GlobeLand30 land cover product in a heterogeneous area, western California. Experimental results represented promises in land cover validation, given that GlobeLand30 showed an overall accuracy of 83.80% with classified samples, which was close to the validation result of 80.45% based on visual interpretation. Additionally, the performances of deep learning based on ResNet-50 and AlexNet were also quantified, revealing no substantial differences in final validation results. The proposed approach ensures geo-tagged photo quality, and supports the sample classification strategy by considering photo distribution, with accuracy improvement from 72.07% to 79.33% compared with solely considering the single nearest photo. Consequently, the presented approach proves the feasibility of deep learning technology on land cover information identification of geo-tagged photos, and has a great potential to support and improve the efficiency of land cover validation. [ABSTRACT FROM AUTHOR]

In recent years, progressive application of convolutional neural networks in image processing has successfully filtered into medical diagnosis. As a prerequisite for images detection and classification, object segmentation in medical images has attracted a great deal of attention. This study is based on the fact that most of the analysis of pathological diagnoses requires nuclei detection as the starting phase for obtaining an insight into the underlying biological process and further diagnosis. In this paper, we introduce an embedded attention model in multi-bridge Wnet (AMB-Wnet) to achieve suppression of irrelevant background areas and obtain good features for learning image semantics and modality to automatically segment nuclei, inspired by the 2018 Data Science Bowl. The proposed architecture, consisting of the redesigned down sample group, up-sample group, and middle block (a new multiple-scale convolutional layers block), is designed to extract different level features. In addition, a connection group is proposed instead of skip-connection to transfer semantic information among different levels. In addition, the attention model is well embedded in the connection group, and the performance of the model is improved without increasing the amount of calculation. To validate the model's performance, we evaluated it using the BBBC038V1 data sets for nuclei segmentation. Our proposed model achieves 85.83% F1-score, 97.81% accuracy, 86.12% recall, and 83.52% intersection over union. The proposed AMB-Wnet exhibits superior results compared to the original U-Net, MultiResUNet, and recent Attention U-Net architecture. [ABSTRACT FROM AUTHOR]