Search

Text clustering is a critical step in text data analysis and has been extensively studied by the text mining community. Most existing text clustering algorithms are based on the bag-of-words model, which faces the high-dimensional and sparsity problems and ignores text structural and sequence information. Deep learning-based models such as convolutional neural networks and recurrent neural networks regard texts as sequences but lack supervised signals and explainable results. In this paper, we propose a deep feature-based text clustering (DFTC) framework that incorporates pretrained text encoders into text clustering tasks. This model, which is based on sequence representations, breaks the dependency on supervision. The experimental results show that our model outperforms classic text clustering algorithms and the state-of-the-art pretrained language model, i.e., BERT, on almost all the considered datasets. In addition, the explanation of the clustering results is significant for understanding the principles of the deep learning approach. Our proposed clustering framework includes an explanation module that can help users understand the meaning and quality of the clustering results.

The Keras deep learning framework is employed to study MRI brain data in a preliminary analysis of brain structure using a convolutional neural network. The results obtained are matched with the content of personality questionnaires. The Big Five personality traits provide easy differentiation for dividing personalities into different groups. Until now, the highest accuracy obtained from the results of personality prediction from the analysis of brain structure is about 70%. Although there is still no effective evidence to prove a clear relationship between brain structure and personality, the obtained results could prove helpful in understanding the basic relationship between brain structure and personality characteristics.

Introduction There are scarce real-world data on the long-term efficacy and safety of cardiopulmonary exercise testing (CPET) combined with the systematic education of cardiac rehabilitation (CR) approach for patients post-coronary stenting, which is, therefore, the subject of this study. Methods Data collected between 1 April 2015 and 20 May 2017 from 11,345 patients in the rehabilitation center database at our hospital were retrospectively analyzed. Five hundred thirty-six patients with incomplete information, or unable to cooperate with telephone follow-up, were excluded; 4001 patients received the combined CR approach; and 6808 patients received only routine post-procedure education (controls). Of these, 2805 CR participants (CR group) were matched 1:1 to controls (control group) using propensity scores. The main outcome was quality of life in Seattle Angina Questionnaire (SAQ) scores. SAQ was measured in hospital and at follow-up; meanwhile, volume/type of habitual exercise, major adverse cardiovascular event (MACE), and its components of target vessel revascularization, myocardial infarction, and cardiac death were recorded and analyzed. Results At median 583 (range 184–963) day follow-up, compared with controls, the CR group showed fewer patients not engaging in physical exercise (22 vs. 956, p

The identification of blood-secretory proteins and the detection of protein biomarkers in the blood have an important clinical application value. Existing methods for predicting blood-secretory proteins are mainly based on traditional machine learning algorithms, and heavily rely on annotated protein features. Unlike traditional machine learning algorithms, deep learning algorithms can automatically learn better feature representations from raw data, and are expected to be more promising to predict blood-secretory proteins. We present a novel deep learning model (DeepHBSP) combined with transfer learning by integrating a binary classification network and a ranking network to identify blood-secretory proteins from the amino acid sequence information alone. The loss function of DeepHBSP in the training step is designed to apply descriptive loss and compactness loss to the binary classification network and the ranking network, respectively. The feature extraction subnetwork of DeepHBSP is composed of a multi-lane capsule network. Additionally, transfer learning is used to train a highly accurate generalized model with small samples of blood-secretory proteins. The main contributions of this study are as follows: 1) a novel deep learning architecture by integrating a binary classification network and a ranking network is proposed, superior to existing traditional machine learning algorithms and other state-of-the-art deep learning architectures for biological sequence analysis; 2) the proposed model for blood-secretory protein prediction uses only amino acid sequences, overcoming the heavy dependence of existing methods on annotated protein features; 3) the blood-secretory proteins predicted by our model are statistically significant compared with existing blood-based biomarkers of cancer.

Background The underlying mechanism of verapamil-sensitive idiopathic left ventricular tachycardia (ILVT) has been postulated to be reentrant activation in the Purkinje fiber network of the left posterior fascicle (LPF) or the left anterior fascicle (LAF). However, changing of cardiac axis deviation in sinus rhythm (SR) or during ILVT after radiofrequency catheter ablation (RFCA) has been rarely analyzed. Methods Of the 232 patients with sustained ILVT induced and surface electrocardiogram (ECG) in SR recorded before and after RFCA, the changes of ECG morphology in SR and during ILVT were analyzed. Results The surface ECG in SR changed in 114 (49.1%) patients after RFCA. ILVT could still be induced in 27 (23.7%) patients. In comparison with the original ILVT, 3 forms of ECG morphology were observed. In 8 patients the ILVT morphology was unchanged. In the 13 patients with ILVT axis deviation conversion after ablation, the successful target was more proximal. In the 6 patients with ILVT morphology change but without axis deviation conversion after ablation, the successful ablation site was more distal. Among 15 patients with recurrent ILVT during follow-up, 7 patients had previous axis deviation changes in SR after RFCA, the changes maintained in 4 patients and recovered in 3 patients. Conclusions The morphology changes on surface ECG in SR after RFCA wouldn't be a necessary prerequisite or a good endpoint for ILVT ablation. To analyze ILVT morphology changes after ablation would help to further clarify an appropriate approach for catheter ablation of ILVT. This article is protected by copyright. All rights reserved.

Aims To establish a simple criterion for determining a failed His-bundle pacing (HBP). This criterion states that if stimulus to QRS end interval is longer than His-bundle potential to QRS end interval (‘S-QRSend > H-QRSend’) then a failed HBP can be determined. Methods and results We performed retrospective analysis on 737 pacing tests around His-bundle in 241 patients and prospective analysis on 400 tests in 123 patients. A successful HBP is defined as that whole His-bundle is captured with or without capture of adjacent ventricular myocardium, otherwise, a failed HBP was considered. The output criteria and effective refractory period criteria were used as the gold standards for determining a successful HBP. The gold standards are that if decreasing the pacing output or pacing cycle length to a certain level results in duration or morphology changes of QRS, then a successful HBP is ascertained. In retrospective analysis of patients with normal His-Purkinje conduction, a failed HBP was determined in 31% (154/492) of pacing tests according to ‘S-QRSend > H-QRSend’; all of them were validated by the gold standards (specificity = 100%). In prospective study, a failed HBP was confirmed according to the simple criterion with 100% accuracy in 33% (79/241) pacing tests. This simple criterion was also suitable for patients with His-Purkinje conduction disease although cases with ‘S-QRSend > H-QRSend’ rarely occurred. Conclusion A failed HBP can be easily and reliably determined solely by ‘S-QRSend > H-QRSend’ in more than 30% pacing tests.

AIM To study the involvement and interrelationship of epithelial cell adhesion molecule (EpCAM) and epithelial-mesenchymal transition (EMT) in endometriosis. METHODS Samples from 114 patients undergoing endometrial biopsy or operation for endometriosis and 23 premenopausal women undergoing endometrial biopsy for non-endometriotic benign disease. Immunohistochemistry was used to detect expression level of EpCAM, E-cadherin and N-cadherin in endometrium from patients with (n = 24) and without endometriosis (n = 23), and in lesions from bowel (n = 46), peritoneal (n = 20) and ovarian (n = 24) endometriosis. RESULTS There was no significant difference in the expression level of EpCAM, E-cadherin and N-cadherin, respectively, between endometrium from women with and without endometriosis (P > 0.05). There was also no significant difference in the expression level of EpCAM, E-cadherin and N-cadherin, respectively, among lesions from the bowel, peritoneal and ovarian endometriosis (P > 0.05). We found that the immunoreactivity of endometriotic epithelial cells to EpCAM and N-cadherin was significantly higher than that of eutopic endometrium, but decreased to E-cadherin (P

A 31-year-old female patient presented with complaints of cyclic low abdominal wall pain and palpable abdominal mass for more than 4 years. Physical examination revealed a painful mass measuring 7 cm × 5 cm within the abdominal wall. Her surgical history included surgery for congenital lymphatic haemangioma twice and caesarean section delivery twice. Results of abdominal wall ultrasonography, magnetic resonance imaging, and enhanced computed tomography revealed a soft tissue mass within the abdominal wall, indicating abdominal wall endometriosis. Pathologic examination with fine-needle aspiration of the abdominal wall mass showed endometrial glands and stroma. A multidisciplinary treatment team was established at admission, and surgical excision of the abdominal wall endometriosis mass was recommended. Surgery was performed by our multidisciplinary treatment team. Intraoperatively, the abdominal wall muscle, symphysis pubis, and anterior bladder wall were found to be infiltrated by abdominal wall endometriosis tissue. The abdominal wall endometriosis mass was completely resected with part of the bladder wall, symphysis pubis periosteum, and abdominal wall muscle and fascia (measuring 9 cm × 8 cm × 6 cm). The abdominal wall defect could not be sutured in a routine manner; thus, autologous reconstruction of the abdominal wall defect with left anterolateral thigh musculocutaneous flap was performed. The patient recovered without complications, and follow-up was uneventful. The successful treatment in our case suggests that adequate preoperative examinations and multidisciplinary treatment team collaboration are crucial to the treatment of patients with large abdominal wall endometriosis mass. Anterolateral thigh musculocutaneous flap reconstruction may serve as an optional treatment for abdominal wall defects during surgical excision of abdominal wall endometriosis mass.

Introduction. Purkinje system and false tendons (FTs) are related to ventricular arrhythmia, but the association between Purkinje fibers and FTs is not clear. This study investigated the associations of anatomical and electrophysiological characteristics between Purkinje fibers and FTs. Methods and Results. We optimized the protocol of Lugol’s iodine solution staining of Purkinje fibers to study the anatomical structure and originated a novel electrophysiological mapping method, named the direct visual mapping (DVM) method, to study the electrophysiological characteristics. By using the above-mentioned innovations in 12 dogs, we found the following. (1) There was no Purkinje fiber found 0.5 cm–1.0 cm below the valve annulus or on the leaflets or chordae tendineae of the mitral valve or adjacent to the top 1/3 of the papillary muscle. (2) Purkinje fibers existed in all FTs, including smaller and tiny FTs. (3) The Purkinje fibers contained in the FTs extended from the proximal to the distal end, and their electrophysiological characteristics were similar to the fibers on the endocardium, including anterograde, retrograde, and decremental conduction and automaticity. Conclusions. Purkinje fibers are commonly found in FTs. The electrophysiological characteristics of the Purkinje fibers contained in FTs are similar to the fibers on the endocardium. FTs might have an anatomical and electrophysiological basis for ventricular arrhythmia.

Automatically describing contents of an image using natural language has drawn much attention because it not only integrates computer vision and natural language processing but also has practical applications. Using an end-to-end approach, we propose a bidirectional semantic attention-based guiding of long short-term memory (Bag-LSTM) model for image captioning. The proposed model consciously refines image features from previously generated text. By fine-tuning the parameters of convolution neural networks, Bag-LSTM obtains more text-related image features via feedback propagation than other models. As opposed to existing guidance-LSTM methods which directly add image features into each unit of an LSTM block, our fine-tuned model dynamically leverages more text-conditional image features, acquired by the semantic attention mechanism, as guidance information. Moreover, we exploit bidirectional gLSTM as the caption generator, which is capable of learning long term relations between visual features and semantic information by making use of both historical and future contextual information. In addition, variations of the Bag-LSTM model are proposed in an effort to sufficiently describe high-level visual-language interactions. Experiments on the Flickr8k and MSCOCO benchmark datasets demonstrate the effectiveness of the model, as compared with the baseline algorithms, such as it is 51.2% higher than BRNN on CIDEr metric.

Accurately identifying protein-ATP (Adenosine-5’-triphosphate) binding sites is significant for protein function annotation and new drug invention. Previous studies often utilize classical machine learning classification algorithms to predict protein-ATP binding sites based on protein primary sequence. However, deep learning as a newly developed technique has shown outstanding performance in various fields. In this work, we introduce the deep convolutional neural network for protein-ATP binding sites prediction based on sequence information. Two classification networks are developed including a residual-inception-based predictor and a multi-inception-based predictor, then the ensemble learning is applied by giving optimized weights to each network architecture to combine them for more superior performance. We examine the performance of our proposed method on two groups of independent testing sets including a classic ATP-binding benchmark dataset and a newly proposed ATP-binding dataset. As a result, our proposed method outperforms other state-of-art sequence-based predictors with the AUC of 0.922 and 0.896 respectively which illustrates the efficacy of deep learning technique in protein-ATP binding sites prediction. The source code and benchmark datasets can be downloaded at https://github.com/tlsjz/ATPbinding .

Urine excretory proteins are among the most commonly used biomarkers in body fluids. Computational identification of urine excretory proteins can provide very useful information for identifying targeted disease biomarkers in urine by linking transcriptome or proteomics data. There are few methods based on conventional machine learning algorithms for predicting urine excretory proteins, and most of these methods strongly depend on the extraction of features from urine excretory proteins. An end-to-end model for urine excretory protein prediction, called DeepUEP, is presented using deep neural networks relying on only amino acid sequence information. The model achieves good performance and outperforms existing methods on training and testing sets. By comparing known urinary protein biomarkers with the results of the model, we find that the model can achieve a true-positive rate of over 80% for urinary protein biomarkers that have been detected in more than one study. We also combine our model with transcriptome and proteomics data from lung cancer patients to predict the potential urinary protein biomarkers of lung cancer. A web server is developed for the prediction of urine excretory proteins, and it can be accessed at the following URL: http://www.csbg-jlu.info/DeepUEP/. We believe that our prediction model and web server are useful for biomedical researchers who are interested in identifying urinary protein biomarkers, especially for candidate proteins in transcriptome or proteomics analyses of diseased tissues.

Measuring conditional relatedness between a pair of genes is a fundamental technique and still a significant challenge in computational biology. Such relatedness can be assessed by gene expression similarities while suffering high false discovery rates. Meanwhile, other types of features, e.g., prior-knowledge based similarities, is only viable for measuring global relatedness. In this paper, we propose a novel machine learning model, named Multi-Features Relatedness (MFR), for accurately measuring conditional relatedness between a pair of genes by incorporating expression similarities with prior-knowledge based similarities in an assessment criterion. MFR is used to predict gene-gene interactions extracted from the COXPRESdb, KEGG, HPRD, and TRRUST databases by the 10-fold cross validation and test verification, and to identify gene-gene interactions collected from the GeneFriends and DIP databases for further verification. The results show that MFR achieves the highest area under curve (AUC) values for identifying gene-gene interactions in the development, test, and DIP datasets. Specifically, it obtains an improvement of 1.1% on average of precision for detecting gene pairs with both high expression similarities and high prior-knowledge based similarities in all datasets, comparing to other linear models and coexpression analysis methods. Regarding cancer gene networks construction and gene function prediction, MFR also obtains the results with more biological significances and higher average prediction accuracy, than other compared models and methods. A website of the MFR model and relevant datasets can be accessed from http://bmbl.sdstate.edu/MFR.

The braking effect of the elevator brake on the elevator is mainly realized by the holding action of the brake shoe on the brake wheel. When the elevator is in normal operation, the brake shoe of the elevator brake will open or brake according to the actual operation of the elevator. With the increase of service life, the relevant parts of the brake may be aging. When the elevator brake shoe is opened, there may still be friction between the brake shoe and the brake wheel, which makes the brake shoe wear and affects the braking performance of the elevator. When the brake shoe rubs against the brake wheel, it will generate heat and make the surface temperature of the brake wheel rise. By monitoring the surface temperature of the brake wheel, the abnormal contact between the brake shoe and the brake wheel can be found. This software takes the elevator brake as the research object, through recording the video of the brake wheel surface during the operation of the elevator captured by the infrared thermal imager, the video of the thermal imager screen is obtained and stored, so as to facilitate the subsequent temperature research and analysis.

Machine learning models can automatically discover biomedical research trends and promote the dissemination of information and knowledge. Text feature representation is a critical and challenging task in natural language processing. Most methods of text feature representation are based on word representation. A good representation can capture semantic and structural information. In this paper, two fusion algorithms are proposed, namely, the Tr-W2v and Ti-W2v algorithms. They are based on the classical text feature representation model and consider the importance of words. The results show that the effectiveness of the two fusion text representation models is better than the classical text representation model, and the results based on the Tr-W2v algorithm are the best. Furthermore, based on the Tr-W2v algorithm, trend analyses of cancer research are conducted, including correlation analysis, keyword trend analysis, and improved keyword trend analysis. The discovery of the research trends and the evolution of hotspots for cancers can help doctors and biological researchers collect information and provide guidance for further research.

On par with aerospace engineering, ocean engineering has caught a lot of attention re-cently. In this paper we employ machine learning and natural language processing methods to reveal new technologies and research hotspots in the ocean engineering field. Our data collection includes 14 high-impact journals, and the abstracts of almost 30,000 papers pub- lished from 2010 to 2019. We employed two topic models, Latent Dirichlet Allocation (LDA) and PhraseLDA. Used independently, the LDA model may lack interpretability and the PhraseLDA result may lose information in the final topics. We hence combined these two models and discovered the research hotspots for each year using affinity propagation cluster- ing and word-cloud-based visualization. The results reveal that several topics such as "wind power" and "ship structure", areas such as the European and Arctic seas, and some common research methods are increasing in popularity. This work consists of data collection, topic modelling, clustering, and visualization, which can help researchers understand the trends and important topics in ocean engineering as well as other fields.

Time Delay Neural Network (TDNN) is a well-performing structure for deep neural network-based speaker recognition systems. In this paper we introduce a novel structure, named Crossed-Time Delay Neural Network (CTDNN) to enhance the performance of current TDNN for speaker recognition. Inspired by the multi-filters setting of convolution layers from convolution neural networks, we set multiple time delay units with different context size at the bottom layer and construct a multilayer parallel network. The proposed CTDNN gives significant improvements over original TDNN on both speaker verification and identification tasks. It outperforms in VoxCeleb1 dataset in verification experiment with a 2.6% absolute Equal Error Rate improvement. In few shots condition, CTDNN reaches 90.4% identification accuracy, which doubles the identification accuracy of original TDNN. We also compare the proposed CTDNN with another new variant of TDNN, Factorized-TDNN, which shows that our model has a 36% absolute identification accuracy improvement under few shots condition. Moreover, the proposed CTDNN can handle training with a larger batch more efficiently and hence, utilize calculation resources economically.

We present brief highlights and updates on some newer projects, both in operation/construction and in preparation stages, of astronomical research on Mainland China, with an emphasis on those involving international collaborations. Limited by the scope of this paper, this sample is not meant to be uniform nor comprehensive, and in some cases it may not be fully up to date. For more specific and detailed information on these or other projects, we refer the readers to the official websites of these projects and those of the National Astronomical Observatories, Chinese Academy of Sciences.

Recent studies uncover that subcellular location of long non-coding RNAs (lncRNAs) can provide significant information on its function. Due to the lack of experimental data, the number of lncRNAs is very limited, experimentally verified subcellular localization, and the numbers of lncRNAs located in different organelle are wildly imbalanced. The prediction of subcellular location of lncRNAs is actually a multi-classification small sample imbalance problem. The imbalance of data results in the poor recognition effect of machine learning models on small data subsets, which is a puzzling and challenging problem in the existing research. In this study, we integrate multi-source features to construct a sequence-based computational tool, lncLocation, to predict the subcellular location of lncRNAs. Autoencoder is used to enhance part of the features, and the binomial distribution-based filtering method and recursive feature elimination (RFE) are used to filter some of the features. It improves the representation ability of data and reduces the problem of unbalanced multi-classification data. By comprehensive experiments on different feature combinations and machine learning models, we select the optimal features and classifier model scheme to construct a subcellular location prediction tool, lncLocation. LncLocation can obtain an 87.78% accuracy using 5-fold cross validation on the benchmark data, which is higher than the state-of-the-art tools, and the classification performance, especially for small class sets, is improved significantly.

Knowledge extracted from the protein-protein interaction (PPI) network can help researchers reveal the molecular mechanisms of biological processes. With the rapid growth in the volume of the biomedical literature, manually detecting and annotating PPIs from raw literature has become increasingly difficult. Hence, automatically extracting PPIs by machine learning methods from raw literature has gained significance in the biomedical research. In this paper, we propose a novel PPI extraction method based on the residual convolutional neural network (CNN). This is the first time that the residual CNN is applied to the PPI extraction task. In addition, the previous state-of-the-art PPI extraction models heavily rely on parsing results from natural language processing tools, such as dependence parsers. Our model does not rely on any parsing tools. We evaluated our model based on five benchmark PPI extraction corpora, AIMed, BioInfer, HPRD50, IEPA, and LLL. The experimental results showed that our model achieved the best results compared with the previous kernel-based and CNN-based PPI extraction models. Compared with the previous recurrent neural network-based PPI extraction models, our model achieved better or comparable performance.

It is always difficult to evacuate crowds in public places like subway stations. The traditional crowd behavior simulation models often ignore two important issues in crowd evacuation: pedestrian tracking and individual differences. To solve the problem, this paper combines social force model (SFM) with deep learning into a novel pedestrian detection method. Firstly, several deep learning algorithms for pedestrian detection were compared, and the best ones for sparse and dense crowds were determined. Next, the pedestrian positions in a real video were acquired by the selected algorithms, and converted into actual coordinates in the scene. Then, the evacuation process was simulated with our method and the SFM based on these coordinates. The results show that our model output closer-to-reality results than the SFM. The research findings shed important new light on evacuation in crowded areas.

Three-phase induction motors (TPIMs) are prone to numerous faults due to their complicated stator and rotor conditions and require a fast response, accurate, and intelligent diagnostic system. Recently developed fault diagnostic systems for induction motors are based on machine learning approaches, but their complex structure typically results in long training time. Moreover, they need to be retrained from scratch if the system is not accurate. We apply incremental broad learning (IBL) method to the diagnosis of TPIM faults. The IBL can train and retrain the network efficiently due to its flexible structure. The new diagnostic framework also consists of feature extraction techniques (empirical mode decomposition and sample entropy) and a non-negative matrix factorization (NMF) IBL approach. The experimental results demonstrate that the IBL system is superior to some algorithms, such as deep belief networks, convolutional neural networks, and extreme learning machine. Moreover, the IBL simplified by NMF is more accurate than the IBL without NMF.

As computer science and information technology are making broad and deep impacts on our daily lives, more and more papers are being submitted to computer science journals and conferences. To help authors decide where they should submit their manuscripts, we present the Content-based Journals & Conferences Recommender System on computer science, as well as its web service at http://www.keaml.cn/prs/ . This system recommends suitable journals or conferences with a priority order based on the abstract of a manuscript. To follow the fast development of computer science and technology, a web crawler is employed to continuously update the training set and the learning model. To achieve interactive online response, we propose an efficient hybrid model based on chi-square feature selection and softmax regression. Our test results show that, the system can achieve an accuracy of 61.37% and suggest the best journals or conferences in about 5 s on average.

Discovering new long non-coding RNAs (lncRNAs) has been a fundamental step in lncRNA-related research. Nowadays, many machine learning-based tools have been developed for lncRNA identification. However, many methods predict lncRNAs using sequence-derived features alone, which tend to display unstable performances on different species. Moreover, the majority of tools cannot be re-trained or tailored by users and neither can the features be customized or integrated to meet researchers’ requirements. In this study, features extracted from sequence-intrinsic composition, secondary structure and physicochemical property are comprehensively reviewed and evaluated. An integrated platform named LncFinder is also developed to enhance the performance and promote the research of lncRNA identification. LncFinder includes a novel lncRNA predictor using the heterologous features we designed. Experimental results show that our method outperforms several state-of-the-art tools on multiple species with more robust and satisfactory results. Researchers can additionally employ LncFinder to extract various classic features, build classifier with numerous machine learning algorithms and evaluate classifier performance effectively and efficiently. LncFinder can reveal the properties of lncRNA and mRNA from various perspectives and further inspire lncRNA–protein interaction prediction and lncRNA evolution analysis. It is anticipated that LncFinder can significantly facilitate lncRNA-related research, especially for the poorly explored species. LncFinder is released as R package (https://CRAN.R-project.org/package=LncFinder). A web server (http://bmbl.sdstate.edu/lncfinder/) is also developed to maximize its availability.

With the massive volume and rapid increasing of data, feature space study is of great importance. To avoid the complex training processes in deep learning models which project original feature space into low-dimensional ones, we propose a novel feature space learning (FSL) model. The main contributions in our approach are: (1) FSL can not only select useful features but also adaptively update feature values and span new feature spaces; (2) four FSL algorithms are proposed with the feature space updating procedure; (3) FSL can provide a better data understanding and learn descriptive and compact feature spaces without the tough training for deep architectures. Experimental results on benchmark data sets demonstrate that FSL-based algorithms performed better than the classical unsupervised, semi-supervised learning and even incremental semi-supervised algorithms. In addition, we show a visualization of the learned feature space results. With the carefully designed learning strategy, FSL dynamically disentangles explanatory factors, depresses the noise accumulation and semantic shift, and constructs easy-to-understand feature spaces.

Motivation Computational methods for phosphorylation site prediction play important roles in protein function studies and experimental design. Most existing methods are based on feature extraction, which may result in incomplete or biased features. Deep learning as the cutting-edge machine learning method has the ability to automatically discover complex representations of phosphorylation patterns from the raw sequences, and hence it provides a powerful tool for improvement of phosphorylation site prediction. Results We present MusiteDeep, the first deep-learning framework for predicting general and kinase-specific phosphorylation sites. MusiteDeep takes raw sequence data as input and uses convolutional neural networks with a novel two-dimensional attention mechanism. It achieves over a 50% relative improvement in the area under the precision-recall curve in general phosphorylation site prediction and obtains competitive results in kinase-specific prediction compared to other well-known tools on the benchmark data. Availability and implementation MusiteDeep is provided as an open-source tool available at https://github.com/duolinwang/MusiteDeep. Supplementary information Supplementary data are available at Bioinformatics online.

Background. Heart failure (HF) is the terminal stage of all cardiovascular events. Although implantable cardioverter defibrillator (ICD) therapies have reduced mortality among the high-risk HF population, it is necessary to determine whether certain factors can predict mortality even after cardiac device implantation. Growth stimulation expressed gene 2 (ST2) is an emerging biomarker for HF patient stratification in different clinical settings. Aims. This study aimed to investigate the relationship between baseline soluble ST2 (sST2) levels in serum and the clinical outcomes of high-risk HF patients with device implantation. Methods. Between January 2017 and August 2018, we prospectively recruited consecutive patients implanted with an ICD for heart failure, with LVEF ≤35% as recommended, and analyzed the basic characteristics, baseline serum sST2, and NT-proBNP levels, with at least 1-year follow-up. All-cause mortality was the primary endpoint. Results. During a 643-day follow-up, all-cause mortality occurred in 16 of 150 patients (10.67%). Incidence of all-cause mortality increased significantly in patients with sST2 levels above 34.98846 ng/ml (16.00% vs. 5.33%, P = 0.034 ). After adjusting the model (age, gender, device implantation, prevention of sudden death, LVEDD, LVEF, WBC and CLBBB, hsTNT, etiology, and eGFR) and the model combined with NT-proBNP, the risk of all-cause death was increased by 2.5% and 1.9%, respectively, per ng/ml of sST2. The best sST2 cutoff for predicting all-cause death was 43.42671 ng/ml (area under the curve: 0.72, sensitive: 0.69, and specificity: 0.69). Compared to patients with sST2 levels below 43.42671 ng/ml, the risk of all-cause mortality was higher in those with values above the threshold (5.1% vs. 21.2%, P = 0.002 ). ST2 level ≥43.42671 ng/ml was an independent predictor of all-cause mortality (HR: 3.30 [95% CI 1.02–10.67]). Age (HR: 1.06 [95% CI: 1.01–1.12]) and increased NT-proBNP per 100 (HR: 1.02 [95% CI: 1.01–1.03]) were also associated with all-cause mortality in ICD patients. Conclusions. sST2 level was associated with risk of all-cause mortality, and a threshold of 43.43 ng/ml showed good distinguishing performance to predict all-cause mortality in patients with severe heart failure, recommended for ICD implantation. Patients with sST2 levels more than 43.42671 ng/ml even after ICD implantation should therefore be monitored carefully.

Traditional Chinese Medicine (TCM) started considering the medicinal and health effects of food thousands of years ago. TCM labels are placed on foods based on cold, neutral, and hot properties similar to Chinese herbal medicine. However, it is unclear whether such a classification has any molecular or biochemical basis, and what the relationship is between this TCM classification and the nutrient composition of food. To answer these questions, we collected a large dataset, in which each type of food has both TCM labels and molecular composition records for statistical analyses and machine-learning predictions. We applied machine-learning methods by using food molecular composition to predict the hot, neutral or cold label of food, and achieved more than 80% accuracy, which clearly indicated that TCM labels have a significant molecular basis. We also applied ANOVA to analyze the main factors contributing to the TCM labels. The ANOVA analysis shows that some molecular/biochemical compositions and categories, such as Energy, Fat, Protein, Water and Selenium (Se), have the strongest correlations with the TCM labels of food. To the best of our knowledge, this study represents the first effort to quantitatively explore the relationship between TCM labels and the molecular composition of food.

Human walking and gaits involve several complex body parts and are influenced by personality, mood, social and cultural traits, and aging. These factors are reflected in shoeprints, which in turn can be used to predict age, a problem not systematically addressed using any computational approach. We collected 100,000 shoeprints of subjects ranging from 7 to 80 years old and used the data to develop a deep learning end-to-end model ShoeNet to analyze age-related patterns and predict age. The model integrates various convolutional neural network models together using a skip mechanism to extract age-related features, especially in pressure and abrasion regions from pair-wise shoeprints. The results show that 40.23% of the subjects had prediction errors within 5-years of age and the prediction accuracy for gender classification reached 86.07%. Interestingly, the age-related features mostly reside in the asymmetric differences between left and right shoeprints. The analysis also reveals interesting age-related and gender-related patterns in the pressure distributions on shoeprints; in particular, the pressure forces spread from the middle of the toe toward outside regions over age with gender-specific variations on heel regions. Such statistics provide insight into new methods for forensic investigations, medical studies of gait-pattern disorders, biometrics, and sport studies., Comment: 24 pages, 20 Figures

Purpose: To evaluate the long-term efficacy of segmental bowel resection for bowel endometriosis and the impact of post-operative complications on clinical outcomes.Methods: 62 symptomatic patients with bowel endometriosis undergoing segmental bowel resection from Jun. 2010 to Jan. 2014 were recruited. A visual analogue scale (VAS) and SF-36 questionnaire were administered before and at least 5 years after surgery. Post-operative complications and pregnancy were also recorded. Median follow-up after operation was 76 months (62-105 months).Results: 62 patients underwent laparoscopic segmental bowel resection, one of which converted to laparotomy. All patients complained of obvious pain symptoms, including dysmenorrhea, dyspareunia, bowel movement pain, chronic pelvic pain and tenesmus. Dysmenorrhea was the most frequent. The relief of all pain symptoms after surgery was statistically significant (P

Breast cancer is estimated to be the leading cancer type among new cases in American women. Core biopsy data have shown a close association between breast hyperplasia and breast cancer. The early diagnosis and treatment of breast hyperplasia are extremely important to prevent breast cancer. The Mongolian medicine RuXian-I is a traditional drug that has achieved a high level of efficacy and a low incidence of side effects in its clinical use. However, for detecting the efficacy of RuXian-I, a rapid and accurate evaluation method based on metabolomic data is still lacking. Therefore, we proposed a framework, named the metabolomics deep belief network (MDBN), to analyze breast hyperplasia metabolomic data. We obtained 168 samples of metabolomic data from an animal model experiment of RuXian-I, which were averaged from control groups, treatment groups, and model groups. In the process of training, unlabelled data were used to pretrain the Deep Belief Networks models, and then labelled data were used to complete fine-tuning based on a limited-memory Broyden Fletcher Goldfarb Shanno (L-BFGS) algorithm. To prevent overfitting, a dropout method was added to the pretraining and fine-tuning procedures. The experimental results showed that the proposed model is superior to other classical classification methods that are based on positive and negative spectra data. Further, the proposed model can be used as an extension of the classification method for metabolomic data. For the high accuracy of classification of the three groups, the model indicates obvious differences and boundaries between the three groups. It can be inferred that the animal model of RuXian-I is well established, which can lay a foundation for subsequent related experiments. This also shows that metabolomic data can be used as a means to verify the effectiveness of RuXian-I in the treatment of breast hyperplasia.

Background His-bundle pacing is currently defined according to the output criteria. However, potential nonselective His-bundle pacing (NSHBP) might be misclassified as right ventricular pacing by the output criteria. Objective The purpose of this study was to use the novel cycle length (CL) criteria by decremental CL pacing to determine the type of pacing and to differentiate NSHBP from right ventricular pacing in particular. Methods His-bundle pacing was performed in 212 patients with normal His-Purkinje conduction (group 1) and 39 patients with His-Purkinje conduction disease for correction of the condition (group 2). The CL criteria state that if decreasing the CL to a certain level results in QRS morphology changes, then NSHBP is ascertained. Results NSHBP was obtained in 170 patients in group 1 and 22 patients in group 2. In group 1, NSHBP was validated in 160 patients by both output criteria and CL criteria. NSHBP was misclassified as right ventricular pacing by the output criteria but was correctly classified by CL criteria in the remaining 10 patients (6%). In group 2, NSHBP was all validated by both criteria. Among the 192 patients with NSHBP, the shortest CL (318 ± 29 ms; range 270–470 ms) with which a stimulus can be conducted along the His bundle was at least 20 ms longer than that of surrounding myocardium (264 ± 16 ms; range 250–330 ms) for each patient, suggesting potentially high sensitivity of the CL criteria. Conclusion CL criteria can determine the types of cardiac pacing independently and can avoid misclassification by output criteria.

Divide and ruling is used to handle classification on complex distributed dataset.AP clustering is better than other clustering methods to divide dataset.An improved PSO and LHS make SA_SVDD parameters self-adaption.Average density is proposed to describe the dataset distribution. The Support Vector Data Description (SVDD) is one of the best-known one-class-classification methods used to solve problems where the sample data are of high dimension but limited amount. The results of SVDD can be greatly affected when the target data are poorly distributed and their density varies extremely. To deal with situations like these, we propose an improved SVDD algorithm, termed as Self-Adaptive SVDD (SA_SVDD), which combines Affinity Propagation (AP) clustering algorithm and SVDD. Firstly, SA_SVDD applies AP clustering to the input data set to obtain a set of compact subclasses, and then the boundary for each subclass is identified by SVDD, and these boundaries are exploited as final classification criteria. AP clustering is tried with different input preferences generated with Latin Hypercube Sampling, and an improved Particle Swarm Optimization (PSO) algorithm evolves the parameters of SVDD. The proposed algorithm is evaluated on a conventional benchmark set and shows significant improvement with respect to other one-class-classification methods. Display Omitted

Background TRIM62 (tripartite motif containing 62) has been found to act as a tumor suppressor of several cancers. However, its precise biological role and related mechanism remain unknown in cervical cancer (CC). Methods Quantitative Real-time PCR and western blot were adopted to detect the mRNA and protein expression level of TRIM62 in both human CC cell lines and tissues. Immunohistochemistry was used to measure the TRIM62 expression in 30 normal cervical and 189 CC tissues. Univariate and multivariate Cox regression analyses and Kaplan–Meier survival analyses performed to investigate the association between TRIM62 expression and CC patients’ prognosis. The effect of TRIM62 on CC growth and metastasis was studied in vitro and in vivo. Multi-pathway reporter array were utilized to identify the potential signaling manipulated by TRIM62. Results TRIM62 was frequently down-regulated in both human CC cells and tissues. Low expression of TRIM62 in CC tissues was associated with aggressive clinicopathological features of CC patients. In addition, TRIM62 was also an independent poor prognostic factor for overall and disease-free survival of CC patients after surgery. Moreover, enforced expression of TRIM62 in CC cells significantly inhibited their abilities of proliferation, migration and invasion in vitro. Besides, subcutaneous xenograft tumor model and xenograft mouse metastatic model respectively displayed that TRIM62 impeded the growth and metastasis of CC in vivo. Furthermore, mechanism study exhibited that TRIM62 could suppress epithelial-mesenchymal transition (EMT) by inhibiting c-Jun/Slug signaling. The inhibitory role of TRIM62 in tumor proliferation might be through regulating cell cycle related proteins CyclinD1 and P27 by targeting c-Jun. Conclusion TRIM62 is a potential prognostic biomarker in CC and suppresses metastasis of CC via inhibiting c-Jun/Slug signaling-mediated EMT. Electronic supplementary material The online version of this article (doi:10.1186/s13046-016-0445-5) contains supplementary material, which is available to authorized users.

In this paper, we propose a novel hybrid text classification model based on deep belief network and softmax regression. To solve the sparse high-dimensional matrix computation problem of texts data, a deep belief network is introduced. After the feature extraction with DBN, softmax regression is employed to classify the text in the learned feature space. In pre-training procedures, the deep belief network and softmax regression are first trained, respectively. Then, in the fine-tuning stage, they are transformed into a coherent whole and the system parameters are optimized with Limited-memory Broyden---Fletcher---Goldfarb---Shanno algorithm. The experimental results on Reuters-21,578 and 20-Newsgroup corpus show that the proposed model can converge at fine-tuning stage and perform significantly better than the classical algorithms, such as SVM and KNN.

The architecture and parameters of convolutional neural networks have an important impact on their performance. To overcome the difficulties of most existing neural architecture search (NAS) methods, including fixed network architecture and huge computing cost, this paper proposes a block-encoding based on neural architecture evolving method. A new block-encoding method is designed to divide the convolutional neural network architecture into blocks consisting of multiple functional layers. Efficient mutating operation is designed to speed up evolutionary search and expand the evolution space of network architecture. Finally, the optimal evolved network is converted into an all-convolutional neural network with fewer parameters and more concise architecture. The experiments on image datasets indicate that the proposed method can greatly reduce network parameters and searching time, achieve competitive classification accuracy and directly obtain the corresponding all-convolutional neural network architecture.

The most common form of an epicardial accessory pathway (AP) has been reported to result from a connection between an extension of the coronary sinus myocardial coat (CSMC) along the middle cardiac vein, posterior coronary vein, or neck of a CS diverticulum [(1)][1]. To our knowledge, only rare

With recent advances in single-cell RNA sequencing, enormous transcriptome datasets have been generated. These datasets have furthered our understanding of cellular heterogeneity and its underlying mechanisms in homogeneous populations. Single-cell RNA sequencing (scRNA-seq) data clustering can group cells belonging to the same cell type based on patterns embedded in gene expression. However, scRNA-seq data are high-dimensional, noisy, and sparse, owing to the limitation of existing scRNA-seq technologies. Traditional clustering methods are not effective and efficient for high-dimensional and sparse matrix computations. Therefore, several dimension reduction methods have been introduced. To validate a reliable and standard research routine, we conducted a comprehensive review and evaluation of four classical dimension reduction methods and five clustering models. Four experiments were progressively performed on two large scRNA-seq datasets using 20 models. Results showed that the feature selection method contributed positively to high-dimensional and sparse scRNA-seq data. Moreover, feature-extraction methods were able to promote clustering performance, although this was not eternally immutable. Independent component analysis (ICA) performed well in those small compressed feature spaces, whereas principal component analysis was steadier than all the other feature-extraction methods. In addition, ICA was not ideal for fuzzy C-means clustering in scRNA-seq data analysis. K-means clustering was combined with feature-extraction methods to achieve good results.

Background Nowadays, because of the huge economic burden on society causing by obesity and diabetes, they turn into the most serious public health challenges in the world. To reveal the close and complex relationships between diabetes, obesity and other diseases, search the effective treatment for them, a novel model named as representative latent Dirichlet allocation (RLDA) topic model is presented. Results RLDA was applied to a corpus of more than 337,000 literatures of diabetes and obesity which were published from 2007 to 2016. To unveil those meaningful relationships between diabetes mellitus, obesity and other diseases, we performed an explicit analysis on the output of our model with a series of visualization tools. Then, with the clinical reports which were not used in the training data to show the credibility of our discoveries, we find that a sufficient number of these records are matched directly. Our results illustrate that in the last 10 years, for obesity accompanying diseases, scientists and researchers mainly focus on 17 of them, such as asthma, gastric disease, heart disease and so on; for the study of diabetes mellitus, it features a more broad scope of 26 diseases, such as Alzheimer’s disease, heart disease and so forth; for both of them, there are 15 accompanying diseases, listed as following: adrenal disease, anxiety, cardiovascular disease, depression, heart disease, hepatitis, hypertension, hypothalamic disease, respiratory disease, myocardial infarction, OSAS, liver disease, lung disease, schizophrenia, tuberculosis. In addition, tumor necrosis factor, tumor, adolescent obesity or diabetes, inflammation, hypertension and cell are going be the hot topics related to diabetes mellitus and obesity in the next few years. Conclusions With the help of RLDA, the hotspots analysis-relation discovery results on diabetes and obesity were achieved. We extracted the significant relationships between them and other diseases such as Alzheimer’s disease, heart disease and tumor. It is believed that the new proposed representation learning algorithm can help biomedical researchers better focus their attention and optimize their research direction. Electronic supplementary material The online version of this article (10.1186/s12918-018-0640-4) contains supplementary material, which is available to authorized users.

Yanchun Liang,1 Duo Liu,1 Fan Yang,1 Wenwei Pan,1 Feitianzhi Zeng,1 Jinjie Wu,2 Hongyu Xie,2 Jiaying Li,2 Shuzhong Yao1 1Department of Obstetrics and Gynecology, The First Affiliated Hospital, Sun Yat-sen University, Guangzhou 510080, China; 2Zhongshan School of Medicine, Sun Yat-sen University, Guangzhou 510080, China Purpose: Recent studies have shown that abnormal distribution of pelvic nerves contributes to endometriosis-associated pain. However, the relationship between neurogenesis and pain severity in endometriosis still remains uncertain, which makes it an enigma for both gynecologists as well as neuropathologists. In this study, we tried to explore a special phenomenon, perineural invasion (PNI), in deep infiltrating endometriosis (DIE) and investigated the correlation between PNI- and DIE-associated pain.Patients and methods: The study was conducted in the Department of Obstetrics and Gynecology of the First Affiliated Hospital of Sun Yat-sen University from June 2012 to January 2015. In total, 64 patients with DIE were enrolled. They received laparoscopically surgical resection of endometriotic lesions. The Kruskal–Wallis and Mann–Whitney tests were used for comparisons of enumeration data. Spearman rank correlation was used for linear analysis.Results: Immunohistochemical analysis demonstrated that PNI was commonly found in DIE lesions. Patients were divided into PNI (+) group and PNI (−) group. The visual analog scale scores of dysmenorrhea, dyspareunia, and chronic pelvic pain were higher in PNI (+) group than in PNI (−) group. Also, we found significantly increased density of newly formed nerve fibers as well as microvessels in lesions of PNI (+) group. Further, double immunofluorescence showed a closely spatial nerve–vessel network in the endometriotic lesion of PNI (+) group. More importantly, correlation analysis revealed positive relation between the density of newly formed nerve fibers in the lesion and the density of microvessels in lesions of PNI (+) group.Conclusion: This study suggests that PNI in endometriotic lesions plays an important role in endometriosis-associated pain, mainly through a mechanism named “neuroangiogenesis”. Keywords: perineural invasion, deep infiltrating endometriosis, pain, neuroangiogenesis

A kind of noncoding RNA with length more than 200 nucleotides named long noncoding RNA (lncRNA) has gained considerable attention in recent decades. Many studies have confirmed that human genome contains many thousands of lncRNAs. LncRNAs play significant roles in many important biological processes, including complex disease diagnosis, prognosis, prevention and treatment. For some important diseases such as cancer, lncRNAs have been novel candidate biomarkers. However, the role of lncRNAs in human diseases is still in its infancy, and only a small part of lncRNA-disease associations have been experimentally verified. Predicting lncRNA-disease association is an important way to understand the mechanism and function of lncRNA involved in diseases to enrich the annotations of lncRNA. Therefore, it is urgent to prioritize lncRNAs potentially associated with diseases. Biological system is a highly complex heterogenous network involved different molecules. Therefore, the algorithms based on network methods have been extensively applied in information fields which can provide a quantifiable characterization for the networks characterizing multifarious biological systems. A heterogeneous network topology possessing abundant interactions between biomedical entities is rarely utilized in similarity-based methods for predicting lncRNA-disease associations based on the array of varying features of lncRNAs and diseases. DeepWalk, encoding the relations of nodes in a continuous vector space, is an extension of language model and unsupervised learning from sequence-based word to network. In this article, we present a novel lncRNA-disease association prediction method based on DeepWalk, which enhances the existing association discovery methods through a topology-based similarity measure. We integrate the heterogeneous data to construct a Linked Tripartite Network which is a heterogeneous network containing three types od nodes which generated from bioinformatics linked datasets and use DeepWalk method to extract topological structure features of the nodes in the linked tripartite network for calculating similarities. Our proposed method can be separated into the following steps: Firstly, we integrate heterogeneous data to construct a Linked Tripartite Network: containing the topological interactions of known lncRNA-disease, lncRNA-microRNA and microRNA-disease. Secondly, the topological structure features of the nodes are extracted based on DeepWalk. Thirdly, similarity scores of disease-disease pairs and lncRNA-lncRNA pairs are computed based on the topology of this network. Finally, new lncRNA and disease associations are discovered by rule-based inference method with lncRNA-lncRNA similarities. Our proposed method shows superior predictive performance for prediction of lncRNA-disease associations based on topological similarity from heterogenous network. The AUC value is used to show the performance of our method. The similarity measurement using network topology based on DeepWalk provide a novel perspective which is different from the similarity derived from sequence or structure information. Availability: All the data and codes are freely availability at: https://github.com/Pengeace/lncRNA-disease-link.

Endometriosis is a complex and heterogeneous disorder with unknown etiology. Dysregulation of macrophages and innervation are important factors influencing the pathogenesis of endometriosis-associated pain. It is known to be an estrogen-dependent disease, estrogen can promote secretion of chemokines from peripheral nerves, enhancing the recruitment and polarization of macrophages in endometriotic tissue. Macrophages have a role in the expression of multiple nerve growth factors (NGF), which mediates the imbalance of neurogenesis in an estrogen-dependent manner. Under the influence of estrogen, co-existence of macrophages and nerves induces an innovative neuro-immune communication. Persistent stimulation by inflammatory cytokines from macrophages on nociceptors of peripheral nerves aggravates neuroinflammation through the release of inflammatory neurotransmitters. This neuro-immune interaction regulated by estrogen sensitizes peripheral nerves, leading to neuropathic pain in endometriosis. The aim of this review is to highlight the significance of estrogen in the interaction between macrophages and nerve fibers, and to suggest a potentially valuable therapeutic target for endometriosis-associated pain.

Endometriosis is a complex and heterogeneous disorder of unknown aetiology. This benign disease possesses special biological behaviours that mimic those of malignant tumours. A pro-endometriotic niche established by an existing lesion is a supportive micro-environment for the progression of endometriosis. After the accumulation of cells by an existing lesion, these components display distinct characteristics that impair immune surveillance. Subsequent retrograde menstruation of endometrial stromal cells into the pro-endometriotic niche facilitates endometriotic progression from early initiation to an advanced lesion. This study aimed to highlight the innovative role of the pro-endometriotic niche in endometriosis, and to provide valuable treatment targets for endometriosis.

A correction to this article has been published and is linked from the HTML and PDF versions of this paper. The error has not been fixed in the paper.

Dimensionality reduction is a common approach to decrease the high computational complexity and improve the classification performance of hyperspectral images. The paper addresses this issue by introducing a novel semisupervised clustering approach to hyperspectral image classification. In this approach, the relationships between the samples (i.e., pixels in hyperspectral data) are measured by two kinds of side constraints, i.e., positive and negative constraints, which are imposed to construct a discriminative transformation that establishes a regularized metric function. Accordingly, a new subspace is built in which the class discrimination capability of each individual feature is expanded, while the spectral correlation among features is greatly reduced. Then, the learned metric is formulated within an exemplar-based clustering framework, i.e., the affinity propagation (AP). Thus, the proposed approach is called decorrelation–separability-based AP (DS-AP). Experimental results obtained on three hyperspectral remote sensing data sets demonstrate the effectiveness of the proposed DS-AP technique for hyperspectral image classification.