BJBN： BERT-JOIN-BiLSTM Networks for Medical Auxiliary Diagnostic

This study proposed a medicine auxiliary diagnosis model based on neural network. The model combines a bidirectional long short-term memory（Bi-LSTM）network and bidirectional encoder representations from transformers (BERT), which can well complete the extraction of local features of Chinese medicine texts. BERT can learn the global information of the text, so use BERT to get the global representation of medical text and then use Bi-LSTM to extract local features. We conducted a large number of comparative experiments on datasets. The results show that the proposed model has significant advantages over the state-of-the-art baseline model. The accuracy of the proposed model is 0.75.

Toward a genome sequence for every animal: Where are we now?

In less than 25 y, the field of animal genome science has transformed from a discipline seeking its first glimpses into genome sequences across the Tree of Life to a global enterprise with ambitions to sequence genomes for all of Earth’s eukaryotic diversity [H. A. Lewin et al., Proc. Natl. Acad. Sci. U.S.A. 115, 4325–4333 (2018)]. As the field rapidly moves forward, it is important to take stock of the progress that has been made to best inform the discipline’s future. In this Perspective, we provide a contemporary, quantitative overview of animal genome sequencing. We identified the best available genome assemblies in GenBank, the world’s most extensive genetic database, for 3,278 unique animal species across 24 phyla. We assessed taxonomic representation, assembly quality, and annotation status for major clades. We show that while tremendous taxonomic progress has occurred, stark disparities in genomic representation exist, highlighted by a systemic overrepresentation of vertebrates and underrepresentation of arthropods. In terms of assembly quality, long-read sequencing has dramatically improved contiguity, whereas gene annotations are available for just 34.3% of taxa. Furthermore, we show that animal genome science has diversified in recent years with an ever-expanding pool of researchers participating. However, the field still appears to be dominated by institutions in the Global North, which have been listed as the submitting institution for 77% of all assemblies. We conclude by offering recommendations for improving genomic resource availability and research value while also broadening global representation.

Speak Global, Sell Local? Digital Linguistic Landscape of Local Small Businesses in the Social Media

This paper focuses on the online presence of languages and linguistic patterns of local small businesses in a bilingual, Hungarian-Romanian ethnic community in Romania. By capturing linguistic diversity and creativity via netnographic research, patterns of linguistic landscape elements in the social media, such as marketing strategy of local small businesses, can be analysed. The findings suggest that despite the need to advertise by using the state language, Romanian, in order to maximize the target audience, the concentration of Hungarian landscape elements is the highest. Businesses construct their linguistic identity by their language choices and practices, aligned with the collective linguistic identity of a bilingual community and the need for a global representation, in order to secure a place in the local market.

Memory-Augmented Transformer for Remote Sensing Image Semantic Segmentation

The semantic segmentation of remote sensing images requires distinguishing local regions of different classes and exploiting a uniform global representation of the same-class instances. Such requirements make it necessary for the segmentation methods to extract discriminative local features between different classes and to explore representative features for all instances of a given class. While common deep convolutional neural networks (DCNNs) can effectively focus on local features, they are limited by their receptive field to obtain consistent global information. In this paper, we propose a memory-augmented transformer (MAT) to effectively model both the local and global information. The feature extraction pipeline of the MAT is split into a memory-based global relationship guidance module and a local feature extraction module. The local feature extraction module mainly consists of a transformer, which is used to extract features from the input images. The global relationship guidance module maintains a memory bank for the consistent encoding of the global information. Global guidance is performed by memory interaction. Bidirectional information flow between the global and local branches is conducted by a memory-query module, as well as a memory-update module, respectively. Experiment results on the ISPRS Potsdam and ISPRS Vaihingen datasets demonstrated that our method can perform competitively with state-of-the-art methods.

Public human microbiome data dominated by highly developed countries

The importance of sampling from globally representative populations has been well established in human genomics. In human microbiome research, however, we lack a full understanding of the global distribution of sampling in research studies. This information is crucial to better understand global patterns of microbiome-associated diseases and to extend the health benefits of this research to all populations. Here, we analyze the country of origin of all 444,829 human microbiome samples that have been collected to date and are available from the world's three largest genomic data repositories, including the Sequence Read Archive (SRA). We show that more than 71% of publicly available human microbiome samples with a known origin come from Europe, the United States, and Canada, including 46.8% from the United States alone, despite the country representing only 4.3% of the global population. We also find that central and southern Asia is the most underrepresented region: Countries such as India, Pakistan, and Bangladesh account for more than a quarter of the world population but make up only 1.8 percent of human microbiome samples. These results demonstrate a critical need to ensure more global representation of participants in microbiome studies.

Multimodel System Identification Based on New Fuzzy Partitioning Similarity Measure

The problem of identifying unstructured nonlinear systems is generally addressed on the basis of multi-model representations involving several linear local models. In the present work, local models are combined to get a global representation using incremental fuzzy clustering. The main contribution is a novel vector similarity measure defined in the System Working Space (SWS) that combines the angular deviation and the usual Euclidean distance. Such a combination makes the new metric highly discriminating leading to a better partitioning of the operating space providing, thereby, a higher accuracy of the model. The developed partitioning method is first evaluated by performing linear local model (LLM) based identification of a academic benchmark multivariable nonlinear system. Then, the performances of the identification method are evaluated using experimental tropospheric ozone data. These evaluations illustrate the supremacy of the new method over the standard Euclidian-distance based partitioning approach.

African Global Representation in Biomedical Sciences

African populations are diverse in their ethnicity, language, culture, and genetics. Although plagued by high disease burdens, until recently the continent has largely been excluded from biomedical studies. Along with limitations in research and clinical infrastructure, human capacity, and funding, this omission has resulted in an underrepresentation of African data and disadvantaged African scientists. This review interrogates the relative abundance of biomedical data from Africa, primarily in genomics and other omics. The visibility of African science through publications is also discussed. A challenge encountered in this review is the relative lack of annotation of data on their geographical or population origin, with African countries represented as a single group. In addition to the abovementioned limitations,the global representation of African data may also be attributed to the hesitation to deposit data in public repositories. Whatever the reason, the disparity should be addressed, as African data have enormous value for scientists in Africa and globally.

Equity in clinical trials for HIV-associated cryptococcal meningitis: A systematic review of global representation and inclusion of patients and researchers

Background It is essential that clinical trial participants are representative of the population under investigation. Using HIV-associated cryptococcal meningitis (CM) as a case study, we conducted a systematic review of clinical trials to determine how inclusive and representative they were both in terms of the affected population and the involvement of local investigators. Methods We searched Medline, EMBASE, Cochrane, Africa-Wide, CINAHL Plus, and Web of Science. Data were extracted for 5 domains: study location and design, screening, participants, researchers, and funders. Data were summarised and compared over 3 time periods: pre-antiretroviral therapy (ART) (pre-2000), early ART (2000 to 2009), and established ART (post-2010) using chi-squared and chi-squared for trend. Comparisons were made with global disease burden estimates and a composite reference derived from observational studies. Results Thirty-nine trials published between 1990 and 2019 were included. Earlier studies were predominantly conducted in high-income countries (HICs) and recent studies in low- and middle-income countries (LMICs). Most recent studies occurred in high CM incidence countries, but some highly affected countries have not hosted trials. The sex and ART status of participants matched those of the general CM population. Patients with reduced consciousness and those suffering a CM relapse were underrepresented. Authorship had poor representation of women (29% of all authors), particularly as first and final authors. Compared to trials conducted in HICs, trials conducted in LMICs were more likely to include female authors (32% versus 20% p = 0.014) but less likely to have authors resident in (75% versus 100%, p < 0.001) or nationals (61% versus 93%, p < 0.001) of the trial location. Conclusions There has been a marked shift in CM trials over the course of the HIV epidemic. Trials are primarily performed in locations and populations that reflect the burden of disease, but severe and relapse cases are underrepresented. Most CM trials now take place in LMICs, but the research is primarily funded and led by individuals and institutions from HICs.

A Trajectory-Based Method for Dynamic Scene Recognition

Existing methods for dynamic scene recognition mostly use global features extracted from the entire video frame or a video segment. In this paper, a trajectory-based dynamic scene recognition method is proposed. A trajectory is formed by a pixel moving across consecutive frames of a video segment. The local regions surrounding the trajectory provide useful appearance and motion information about a portion of the video segment. The proposed method works at several stages. First, dense and evenly distributed trajectories are extracted from a video segment. Then, the fully-connected-layer features are extracted from each trajectory using a pre-trained Convolutional Neural Networks (CNNs) model, forming a feature sequence. Next, these feature sequences are fed into a Long-Short-Term-Memory (LSTM) network to learn their temporal behavior. Finally, by aggregating the information of the trajectories, a global representation of the video segment can be obtained for classification purposes. The LSTM is trained using synthetic trajectory feature sequences instead of real ones. The synthetic feature sequences are generated with a series of generative adversarial networks (GANs). In addition to classification, category-specific discriminative trajectories are located in a video segment, which help reveal what portions of a video segment are more important than others. This is achieved by formulating an optimization problem to learn discriminative part detectors for all categories simultaneously. Experimental results on two benchmark dynamic scene datasets show that the proposed method is very competitive with six other methods.