UPANets: Learning from the Universal Pixel Attention Neworks.

With the successful development in computer vision, building a deep convolutional neural network (CNNs) has been mainstream, considering the character of shared parameters in a convolutional layer. Stacking convolutional layers into a deep structure improves performance, but over-stacking also ramps up the needed resources for GPUs. Seeing another surge of Transformers in computer vision, the issue has aroused severely. A resource-hungry model is hardly implemented for limited hardware or single-customers-based GPU. Therefore, this work focuses on these concerns and proposes an efficient but robust backbone, which equips with channel and spatial direction attentions, so the attentions help to expand receptive fields in shallow convolutional layers and pass the information to every layer. An attention-boosted network based on already efficient CNNs, Universal Pixel Attention Networks (UPANets), is proposed. Through a series of experiments, UPANets fulfil the purposes of learning global information with less needed resources and outshine many existing SOTAs in CIFAR-{10, 100}.

Random RotBoost: An Ensemble Classification Method Based on Rotation Forest and AdaBoost in Random Subsets and Its Application to Clinical Decision Support.

In the era of bathing in big data, it is common to see enormous amounts of data generated daily. As for the medical industry, not only could we collect a large amount of data, but also see each data set with a great number of features. When the number of features is ramping up, a common dilemma is adding computational cost during inferring. To address this concern, the data rotational method by PCA in tree-based methods shows a path. This work tries to enhance this path by proposing an ensemble classification method with an AdaBoost mechanism in random, automatically generating rotation subsets termed Random RotBoost. The random rotation process has replaced the manual pre-defined number of subset features (free pre-defined process). Therefore, with the ensemble of the multiple AdaBoost-based classifier, overfitting problems can be avoided, thus reinforcing the robustness. In our experiments with real-world medical data sets, Random RotBoost reaches better classification performance when compared with existing methods. Thus, with the help from our proposed method, the quality of clinical decisions can potentially be enhanced and supported in medical tasks.

Regional Population Forecast and Analysis Based on Machine Learning Strategy.

Regional population forecast and analysis is of essence to urban and regional planning, and a well-designed plan can effectively construct a sound national infrastructure and stabilize positive population growth. Traditionally, either urban or regional planning relies on the opinions of demographers in terms of how the population of a city or a region will grow. Multi-regional population forecast is currently possible, carried out mainly on the basis of the Interregional Cohort-Component model. While this model has its unique advantages, several demographic rates are determined based on the decisions made by primary planners. Hence, the only drawback for cohort-component type population forecasting is allowing the analyst to specify the demographic rates of the future, and it goes without saying that this tends to introduce a biased result in forecasting accuracy. To effectively avoid this problem, this work proposes a machine learning-based method to forecast multi-regional population growth objectively. Thus, this work, drawing upon the newly developed machine learning technology, attempts to analyze and forecast the population growth of major cities in Taiwan. By effectively using the advantage of the XGBoost algorithm, the evaluation of feature importance and the forecast of multi-regional population growth between the present and the near future can be observed objectively, and it can further provide an objective reference to the urban planning of regional population.

A novel bagging C4.5 algorithm based on wrapper feature selection for supporting wise clinical decision making.

From the perspective of clinical decision-making in a Medical IoT-based healthcare system, achieving effective and efficient analysis of long-term health data for supporting wise clinical decision-making is an extremely important objective, but determining how to effectively deal with the multi-dimensionality and high volume of generated data obtained from Medical IoT-based healthcare systems is an issue of increasing importance in IoT healthcare data exploration and management. A novel classifier or predicator equipped with a good feature selection function contributes effectively to classification and prediction performance. This paper proposes a novel bagging C4.5 algorithm based on wrapper feature selection, for the purpose of supporting wise clinical decision-making in the medical and healthcare fields. In particular, the new proposed sampling method, S-C4.5-SMOTE, is not only able to overcome the problem of data distortion, but also improves overall system performance because its mechanism aims at effectively reducing the data size without distortion, by keeping datasets balanced and technically smooth. This achievement directly supports the Wrapper method of effective feature selection without the need to consider the problem of huge amounts of data; this is a novel innovation in this work.

Real-time Automatic M-mode Echocardiography Measurement with Panel Attention.

Motion mode (M-mode) echocardiography is essential for measuring cardiac dimension and ejection fraction. However, the current diagnosis is time-consuming and suffers from diagnosis accuracy variance. This work resorts to building an automatic scheme through well-designed and well-trained deep learning to conquer the situation. That is, we proposed RAMEM, an automatic scheme of real-time M-mode echocardiography, which contributes three aspects to address the challenges: 1) provide MEIS, the first dataset of M-mode echocardiograms, to enable consistent results and support developing an automatic scheme; For detecting objects accurately in echocardiograms, it requires big receptive field for covering long-range diastole to systole cycle. However, the limited receptive field in the typical backbone of convolutional neural networks (CNN) and the losing information risk in non-local block (NL) equipped CNN risk the accuracy requirement. Therefore, we 2) propose panel attention embedding with updated UPANets V2, a convolutional backbone network, in a real-time instance segmentation (RIS) scheme for boosting big object detection performance; 3) introduce AMEM, an efficient algorithm of automatic M-mode echocardiography measurement, for automatic diagnosis; The experimental results show that RAMEM surpasses existing RIS schemes (CNNs with NL & Transformers as the backbone) in PASCAL 2012 SBD and human performances in MEIS. The implemented code and dataset are available at https://github.com/hanktseng131415go/RAMEM.

Distinct effects of ruxolitinib and interferon-alpha on murine JAK2V617F myeloproliferative neoplasm hematopoietic stem cell populations.

JAK2V617F is the most common mutation in patients with BCR-ABL negative myeloproliferative neoplasms (MPNs). The eradication of JAK2V617F hematopoietic stem cells (HSCs) is critical for achieving molecular remissions and cure. We investigate the distinct effects of two therapies, ruxolitinib (JAK1/2 inhibitor) and interferon-alpha (IFN-α), on the disease-initiating HSC population. Whereas ruxolitinib inhibits Stat5 activation in erythroid progenitor populations, it fails to inhibit this same pathway in HSCs. In contrast, IFN-α has direct effects on HSCs. Furthermore, STAT1 phosphorylation and pathway activation is greater after IFN-α stimulation in Jak2V617F murine HSCs with increased induction of reactive oxygen species, DNA damage and reduction in quiescence after chronic IFN-α treatment. Interestingly, ruxolitinib does not block IFN-α induced reactive oxygen species and DNA damage in Jak2V617F murine HSCs in vivo. This work provides a mechanistic rationale informing how pegylated IFN-α reduces JAK2V617F allelic burden in the clinical setting and may inform future clinical efforts to combine ruxolitinib with pegylated IFN-α in patients with MPN.

Protein secondary structure prediction: A survey of the state of the art.

Protein secondary structure prediction (PSSP) is a fundamental task in protein science and computational biology, and it can be used to understand protein 3-dimensional (3-D) structures, further, to learn their biological functions. In the past decade, a large number of methods have been proposed for PSSP. In order to learn the latest progress of PSSP, this paper provides a survey on the development of this field. It first introduces the background and related knowledge of PSSP, including basic concepts, data sets, input data features and prediction accuracy assessment. Then, it reviews the recent algorithmic developments of PSSP, which mainly focus on the latest decade. Finally, it summarizes the corresponding tendencies and challenges in this field. This survey concludes that although various PSSP methods have been proposed, there still exist several further improvements or potential research directions. We hope that the presented guidelines will help nonspecialists and specialists to learn the critical progress in PSSP in recent years.

Similarity/dissimilarity calculation methods of DNA sequences: A survey.

DNA sequence similarity/dissimilarity analysis is a fundamental task in computational biology, which is used to analyze the similarity of different DNA sequences for learning their evolutionary relationships. In past decades, a large number of similarity analysis methods for DNA sequence have been proposed due to the ever-growing demands. In order to learn the advances of DNA sequence similarity analysis, we make a survey and try to promote the development of this field. In this paper, we first introduce the related knowledge of DNA similarities analysis, including the data sets, similarities distance and output data. Then, we review recent algorithmic developments for DNA similarity analysis to represent a survey of the art in this field. At last, we summarize the corresponding tendencies and challenges in this research field. This survey concludes that although various DNA similarity analysis methods have been proposed, there still exist several further improvements or potential research directions in this field.

The dimer interface of the SARS coronavirus nucleocapsid protein adapts a porcine respiratory and reproductive syndrome virus-like structure.

We have employed NMR to investigate the structure of SARS coronavirus nucleocapsid protein dimer. We found that the secondary structure of the dimerization domain consists of five alpha helices and a beta-hairpin. The dimer interface consists of a continuous four-stranded beta-sheet superposed by two long alpha helices, reminiscent of that found in the nucleocapsid protein of porcine respiratory and reproductive syndrome virus. Extensive hydrogen bond formation between the two hairpins and hydrophobic interactions between the beta-sheet and the alpha helices render the interface highly stable. Sequence alignment suggests that other coronavirus may share the same structural topology.

Structure of the subunit binding domain and dynamics of the di-domain region from the core of human branched chain alpha-ketoacid dehydrogenase complex.

The homo-24-meric dihydrolipoyl transacylase (E2) scaffold of the human branched-chain alpha-ketoacid dehydrogenase complex (BCKDC) contains the lipoyl-bearing domain (hbLBD), the subunit-binding domain (hbSBD) and the inner core domain that are linked to carry out E2 functions in substrate channeling and recognition. In this study, we employed NMR techniques to determine the structure of hbSBD and dynamics of several truncated constructs from the E2 component of the human BCKDC, including hbLBD (residues 1-84), hbSBD (residues 111-149), and a di-domain (hbDD) (residues 1-166) comprising hbLBD, hbSBD and the interdomain linker. The solution structure of hbSBD consists of two nearly parallel helices separated by a long loop, similar to the structures of the SBD isolated from other species, but it lacks the short 3(10) helix. The NMR results show that the structures of hbLBD and hbSBD in isolated forms are not altered by the presence of the interdomain linker in hbDD. The linker region is not entirely exposed to solvent, where amide resonances associated with approximately 50% of the residues are observable. However, the tethering of these two domains in hbDD significantly retards the overall rotational correlation times of hbLBD and hbSBD, changing from 5.54 ns and 5.73 ns in isolated forms to 8.37 ns and 8.85 ns in the linked hbDD, respectively. We conclude that the presence of the interdomain linker restricts the motional freedom of the hbSBD more significantly than hbLBD, and that the linker region likely exists as a soft rod rather than a flexible string in solution.

Modular organization of SARS coronavirus nucleocapsid protein.

The SARS-CoV nucleocapsid (N) protein is a major antigen in severe acute respiratory syndrome. It binds to the viral RNA genome and forms the ribonucleoprotein core. The SARS-CoV N protein has also been suggested to be involved in other important functions in the viral life cycle. Here we show that the N protein consists of two non-interacting structural domains, the N-terminal RNA-binding domain (RBD) (residues 45-181) and the C-terminal dimerization domain (residues 248-365) (DD), surrounded by flexible linkers. The C-terminal domain exists exclusively as a dimer in solution. The flexible linkers are intrinsically disordered and represent potential interaction sites with other protein and protein-RNA partners. Bioinformatics reveal that other coronavirus N proteins could share the same modular organization. This study provides information on the domain structure partition of SARS-CoV N protein and insights into the differing roles of structured and disordered regions in coronavirus nucleocapsid proteins.

Sequential structural changes of Escherichia coli thioesterase/protease I in the serial formation of Michaelis and tetrahedral complexes with diethyl p-nitrophenyl phosphate.

Escherichia coli thioesterase/protease I (TEP-I) belongs to a new subclass of lipolytic enzymes of the serine hydrolase superfamily. Here we report the first direct NMR observation of the formation of the Michaelis complex (MC) between TEP-I and diethyl p-nitrophenyl phosphate (DENP), an active site directed inhibitor of serine protease, and its subsequent conversion to the tetrahedral complex (TC). NMR, ESI-MS, and kinetic data showed that DENP binds to TEP-I in a two-step process, a fast formation of MC followed by a slow conversion to TC. NMR chemical shift perturbation further revealed that perturbations were confined mainly to four conserved segments comprising the active site. Comparable magnitudes of chemical shift perturbations were detected in both steps. The largest chemical shift perturbation occurred around the catalytic Ser(10). In MC, the conformation of the mobile Ser(10) was stabilized, and its amide resonance became observable. From the large chemical shift perturbation upon conversion from MC to TC, we propose that the amide protons of Ser(10) and Gly(44) serve as the oxyanion hole proton donors that stabilize the tetrahedral adduct. The pattern of residues perturbed in both steps suggests a sequential, stepwise structural change upon binding of DENP. The present study also demonstrates the important catalytic roles of conserved residues in the SGNH family of proteins.