Search

Federated learning (FL) is a method used for training artificial intelligence models with data from multiple sources while maintaining data anonymity, thus removing many barriers to data sharing. Here we used data from 20 institutes across the globe to train a FL model, called EXAM (electronic medical record (EMR) chest X-ray AI model), that predicts the future oxygen requirements of symptomatic patients with COVID-19 using inputs of vital signs, laboratory data and chest X-rays. EXAM achieved an average area under the curve (AUC) >0.92 for predicting outcomes at 24 and 72 h from the time of initial presentation to the emergency room, and it provided 16% improvement in average AUC measured across all participating sites and an average increase in generalizability of 38% when compared with models trained at a single site using that site’s data. For prediction of mechanical ventilation treatment or death at 24 h at the largest independent test site, EXAM achieved a sensitivity of 0.950 and specificity of 0.882. In this study, FL facilitated rapid data science collaboration without data exchange and generated a model that generalized across heterogeneous, unharmonized datasets for prediction of clinical outcomes in patients with COVID-19, setting the stage for the broader use of FL in healthcare. Federated learning, a method for training artificial intelligence algorithms that protects data privacy, was used to predict future oxygen requirements of symptomatic patients with COVID-19 using data from 20 different institutes across the globe.

Given that mass gathering events involve heterogeneous and time-varying contact between residents and visitors, we sought to identify possible measures to prevent the potential acceleration of the outbreak of an emerging infectious disease induced by such events. An individual-based simulator was built based on a description of the reproduction rate among people infected with the infectious disease in a hypothetical city. Three different scenarios were assessed using our simulator, in which controls aimed at reduced contact were assumed to be carried out only in the main event venue or at subsequent additional events, or in which behavior restrictions were carried out among the visitors to the main event. The simulation results indicated that the increase in the number of patients with COVID-19 could possibly be suppressed to a level equivalent to that if the event were not being held so long as the prevalence among visitors was only slightly higher than that among domestic residents and strict requirements were applied to the activities of visitors.

Microwave (MW) together with prior water presoaking was used as a pretreatment for enzymatic retting of flax stems in this work. Weight loss, pectin removal, the resulting fibers’ properties and the degree of water absorption of the stems after MW pretreatment were measured and analyzed. The effect of MW pretreatment on enzymatic retting was determined and compared with the radio frequency (RF) pretreatment which adopted the same processing parameters. It was found that the 8 h-30 min MW pretreatment improved the enzymatic retting efficiency most, increasing weight loss by 39 %, decreasing pectin content by 10.96 % compared to the control enzymatic retting and obtaining water sorption of 2.23, and produced fibers with reasonable tensile performance. Compared with RF pretreatment, MW pretreatment enabled flax to absorb more water and more energy then got more improvement in the retting efficiency. And the 30 min-MW treatment could improve pectin removal for the non-presoaked flax as compared to the control enzymatic retting, which was different from RF treatment. The enzyme-retted fibers with the same MW and RF pretreatments had close tensile properties, however, their values of the MW-pretreated fibers tended to be slightly lower. It was also found that the tensile performance of the pretreated, enzyme-retted fibers was linked to the amount of energy absorbed, the temperature-maintaining time proportion during MW or RF treatment, and whether the flax had been presoaked or not.

Radio frequency (RF) pretreatment of different water-presoaking time and RF treatment durations was introduced into enzymatic retting. RF pretreatment was used to degrade non-cellulosic materials and encourage the fast water-absorbing and swelling of the polar components in flax phloem, thus loosen the interfaces between fibers and fiber bundles and facilitate enzymes penetration. Retting efficiency, the resulting fibers properties and the water sorption after RF pretreatment were measured. It was found that the 8h-30min RF pretreatment improved retting efficiency most, increasing weight loss by 47 % and decreasing pectin content by 7.87 % compared to the control enzymatic retting and obtaining water sorption of 2.19. The results RF treated presoaked and non-presoaked flax were found to be different. The RF treatment of 30 min for the non-presoaked stems did not cause any improvement in pectin removal with a percentage decrease in pectin content of -8.41 %. The reason probably was that partial pectic substances in the non-presoaked flax phloem were dried, hardened and compacted during RF treatment. It was also found that RF pretreatment improved fibers whiteness and tended to weaken fibers tensile performance, and how it affected the fibers tensile properties was closely linked to the RF pretreatment processing.

Cancer subtype identification is an unmet need in precision diagnosis. Recently, evolutionary conservation has been indicated containing understandable signatures for functional significance in cancers. However, the importance of evolutionary conservation in distinguishing cancer subtypes remains unclear. Here, we identified the evolutionarily conserved genes (i.e., core gene) and observed that they are mainly involved in the pathways relevant to cell growth and metabolisms. By using these core genes, we integrated their evolutionary and genomic profiles with deep learning to develop a feature-based strategy (FES) and an image-based strategy (IMS). In comparison with FES using the random set and the strategy using the PAM50 classifier, core gene set-based FES has higher accuracy for identifying breast cancer subtypes. Moreover, the IMS with data augmentation yields better performance than the other strategies. Comprehensive analysis of eight TCGA cancer data demonstrates that our evolutionary conservation-based models provide a valid and helpful approach to identify cancer subtypes and the core gene set offers distinguishable clues of cancer subtypes.

It is known that many driver nodes are required to control complex biological networks. Previous studies imply that O(N) driver nodes are required in both linear complex network and Boolean network models with N nodes if an arbitrary state is specified as the target. In this paper, we mathematically prove under a reasonable assumption that the expected number of driver nodes is only O(log2N + log2M) for controlling Boolean networks if the targets are restricted to attractors, where M is the number of attractors. Since it is expected that M is not very large in many practical networks, this is a significant improvement. This result is based on discovery of novel relationships between control problems on Boolean networks and the coupon collector’s problem, a well-known concept in combinatorics. We also provide lower bounds of the number of driver nodes as well as simulation results using artificial and realistic network data, which support our theoretical findings.

It is known that many driver nodes are required to control complex biological networks. Previous studies imply that O(N) driver nodes are required in both linear complex network and Boolean network models with N nodes if an arbitrary state is specified as the target. In order to cope with this intrinsic difficulty, we consider a special case of the control problem in which the targets are restricted to attractors. For this special case, we mathematically prove under the uniform distribution of states in basins that the expected number of driver nodes is only O ( log 2 N + log 2 M ) for controlling Boolean networks, where M is the number of attractors. Since it is expected that M is not very large in many practical networks, the new model requires a much smaller number of driver nodes. This result is based on discovery of novel relationships between control problems on Boolean networks and the coupon collector’s problem, a well-known concept in combinatorics. We also provide lower bounds of the number of driver nodes as well as simulation results using artificial and realistic network data, which support our theoretical findings.

Background Since many proteins become functional only after they interact with their partner proteins and form protein complexes, it is essential to identify the sets of proteins that form complexes. Therefore, several computational methods have been proposed to predict complexes from the topology and structure of experimental protein-protein interaction (PPI) network. These methods work well to predict complexes involving at least three proteins, but generally fail at identifying complexes involving only two different proteins, called heterodimeric complexes or heterodimers. There is however an urgent need for efficient methods to predict heterodimers, since the majority of known protein complexes are precisely heterodimers. Results In this paper, we use three promising kernel functions, Min kernel and two pairwise kernels, which are Metric Learning Pairwise Kernel (MLPK) and Tensor Product Pairwise Kernel (TPPK). We also consider the normalization forms of Min kernel. Then, we combine Min kernel or its normalization form and one of the pairwise kernels by plugging. We applied kernels based on PPI, domain, phylogenetic profile, and subcellular localization properties to predicting heterodimers. Then, we evaluate our method by employing C-Support Vector Classification (C-SVC), carrying out 10-fold cross-validation, and calculating the average F-measures. The results suggest that the combination of normalized-Min-kernel and MLPK leads to the best F-measure and improved the performance of our previous work, which had been the best existing method so far. Conclusions We propose new methods to predict heterodimers, using a machine learning-based approach. We train a support vector machine (SVM) to discriminate interacting vs non-interacting protein pairs, based on informations extracted from PPI, domain, phylogenetic profiles and subcellular localization. We evaluate in detail new kernel functions to encode these data, and report prediction performance that outperforms the state-of-the-art.

Water retting is a traditional retting method that enables the low-cost production of bast fibers. This study investigated the retting of flax straws by distilled water for three different durations at room temperature in laboratory condition. The retting quality was evaluated in terms of the weight loss and degumming rate together with the fiber properties, which included color, linear density, and tensile properties. The degumming rate was defined as the percentage change in pectin content of phloem regions from the raw flax to water-retted flax. It was found that the dissolution of pectin and other contaminating materials during the beginning retting stage must have played an important role in pectin (content) and weight loss besides pectin degradation, and water retting gradually improved both the apparel properties, such as whiteness and fineness, and the mechanical properties of the fibers. Given the results, a water retting duration of six days should be sufficient to provide sound retting efficiency and reasonable fiber properties.

In this paper, we study domain compositions of proteins via compression of whole proteins in an organism for the sake of obtaining the entropy that the individual contains. We suppose that a protein is a multiset of domains. Since gene duplication and fusion have occurred through evolutionary processes, the same domains and the same compositions of domains appear in multiple proteins, which enables us to compress a proteome by using references to proteins for duplicated and fused proteins. Such a network with references to at most two proteins is modeled as a directed hypergraph. We propose a heuristic approach by combining the Edmonds algorithm and an integer linear programming, and apply our procedure to 14 proteomes of Dictyostelium discoideum, Escherichia coli, Saccharomyces cerevisiae, Schizosaccharomyces pombe, Caenorhabditis elegans, Drosophila melanogaster, Arabidopsis thaliana, Oryza sativa, Danio rerio, Xenopus laevis, Gallus gallus, Mus musculus, Pan troglodytes, and Homo sapiens. The compressed size using both of duplication and fusion was smaller than that using only duplication, which suggests the importance of fusion events in evolution of a proteome.

Cancer subtype identification is an unmet need in precision diagnosis. Recently, evolutionary conservation has been indicated to contain informative signatures for functional significance in cancers. However, the importance of evolutionary conservation in distinguishing cancer subtypes remains largely unclear. Here, we identified the evolutionarily conserved genes (i.e. core genes) and observed that they are primarily involved in cellular pathways relevant to cell growth and metabolisms. By using these core genes, we developed two novel strategies, namely a feature-based strategy (FES) and an image-based strategy (IMS) by integrating their evolutionary and genomic profiles with the deep learning algorithm. In comparison with the FES using the random set and the strategy using the PAM50 classifier, the core gene set-based FES achieved a higher accuracy for identifying breast cancer subtypes. The IMS and FES using the core gene set yielded better performances than the other strategies, in terms of classifying both breast cancer subtypes and multiple cancer types. Moreover, the IMS is reproducible even using different gene expression data (i.e. RNA-seq and microarray). Comprehensive analysis of eight cancer types demonstrates that our evolutionary conservation-based models represent a valid and helpful approach for identifying cancer subtypes and the core gene set offers distinguishable clues of cancer subtypes.

Grammar-based compression is to find a small grammar that generates a given data and has been well-studied in text compression. In this paper, we apply this methodology to compression of rectangular image data. We first define a context-free rectangular image grammar CFRIG by extending the context-free grammar. Then we propose a quadsection type algorithm by extending a bisection type algorithm for grammar-based compression of text data. We show that our proposed algorithm approximates in polynomial time the smallest CFRIG within a factor of On^{4/3}, where an input image data is of size On × On. Furthermore, we practically improve the proposed algorithm by adding several rules concerning sub-images. We also present results on computational experiments on the proposed algorithms.

[Background]Protein complexes play important roles in biological systems such as gene regulatory networks and metabolic pathways. Most methods for predicting protein complexes try to find protein complexes with size more than three. It, however, is known that protein complexes with smaller sizes occupy a large part of whole complexes for several species. In our previous work, we developed a method with several feature space mappings and the domain composition kernel for prediction of heterodimeric protein complexes, which outperforms existing methods. [Results]We propose methods for prediction of heterotrimeric protein complexes by extending techniques in the previous work on the basis of the idea that most heterotrimeric protein complexes are not likely to share the same protein with each other. We make use of the discriminant function in support vector machines (SVMs), and design novel feature space mappings for the second phase. As the second classifier, we examine SVMs and relevance vector machines (RVMs). We perform 10-fold cross-validation computational experiments. The results suggest that our proposed two-phase methods and SVM with the extended features outperform the existing method NWE, which was reported to outperform other existing methods such as MCL, MCODE, DPClus, CMC, COACH, RRW, and PPSampler for prediction of heterotrimeric protein complexes. [Conclusions]We propose two-phase prediction methods with the extended features, the domain composition kernel, SVMs and RVMs. The two-phase method with the extended features and the domain composition kernel using SVM as the second classifier is particularly useful for prediction of heterotrimeric protein complexes.

Background Modeling high-dimensional data involving thousands of variables is particularly important for gene expression profiling experiments, nevertheless,it remains a challenging task. One of the challenges is to implement an effective method for selecting a small set of relevant genes, buried in high-dimensional irrelevant noises. RELIEF is a popular and widely used approach for feature selection owing to its low computational cost and high accuracy. However, RELIEF based methods suffer from instability, especially in the presence of noisy and/or high-dimensional outliers. Results We propose an innovative feature weighting algorithm, called LHR, to select informative genes from highly noisy data. LHR is based on RELIEF for feature weighting using classical margin maximization. The key idea of LHR is to estimate the feature weights through local approximation rather than global measurement, which is typically used in existing methods. The weights obtained by our method are very robust in terms of degradation of noisy features, even those with vast dimensions. To demonstrate the performance of our method, extensive experiments involving classification tests have been carried out on both synthetic and real microarray benchmark datasets by combining the proposed technique with standard classifiers, including the support vector machine (SVM), k-nearest neighbor (KNN), hyperplane k-nearest neighbor (HKNN), linear discriminant analysis (LDA) and naive Bayes (NB). Conclusion Experiments on both synthetic and real-world datasets demonstrate the superior performance of the proposed feature selection method combined with supervised learning in three aspects: 1) high classification accuracy, 2) excellent robustness to noise and 3) good stability using to various classification algorithms.

Since many proteins express their functional activity by interacting with other proteins and forming protein complexes, it is very useful to identify sets of proteins that form complexes. For that purpose, many prediction methods for protein complexes from protein-protein interactions have been developed such as MCL, MCODE, RNSC, PCP, RRW, and NWE. These methods have dealt with only complexes with size of more than three because the methods often are based on some density of subgraphs. However, heterodimeric protein complexes that consist of two distinct proteins occupy a large part according to several comprehensive databases of known complexes. In this paper, we propose several feature space mappings from protein-protein interaction data, in which each interaction is weighted based on reliability. Furthermore, we make use of prior knowledge on protein domains to develop feature space mappings, domain composition kernel and its combination kernel with our proposed features. We perform ten-fold cross-validation computational experiments. These results suggest that our proposed kernel considerably outperforms the naive Bayes-based method, which is the best existing method for predicting heterodimeric protein complexes.

Grammar-based compression is to find a small grammar that generates a given data and has been well-studied in text compression. In this paper, we apply this methodology to compression of rectangular image data. We first define a context-free rectangular image grammar (CFRIG) by extending the context-free grammar. Then we propose a quadsection type algorithm by extending a bisection type algorithm for grammar-based compression of text data. We show that our proposed algorithm approximates in polynomial time the smallest CFRIG within a factor of O(n 4/3), where an input image data is of size O(n) ×O(n). We also present results on computational experiments on the proposed algorithm.