Search

Owing the epidemic of the novel coronavirus disease 2019 (COVID-19), chest X-ray computed tomography imaging is being used for effectively screening COVID-19 patients. The development of computer-aided systems based on deep neural networks (DNNs) has become an advanced open source to rapidly and accurately detect COVID-19 cases because the need for expert radiologists, who are limited in number, forms a bottleneck for screening. However, thus far, the vulnerability of DNN-based systems has been poorly evaluated, although realistic and high-risk attacks using universal adversarial perturbation (UAP), a single (input image agnostic) perturbation that can induce DNN failure in most classification tasks, are available. Thus, we focus on representative DNN models for detecting COVID-19 cases from chest X-ray images and evaluate their vulnerability to UAPs. We consider non-targeted UAPs, which cause a task failure, resulting in an input being assigned an incorrect label, and targeted UAPs, which cause the DNN to classify an input into a specific class. The results demonstrate that the models are vulnerable to non-targeted and targeted UAPs, even in the case of small UAPs. In particular, the 2% norm of the UAPs to the average norm of an image in the image dataset achieves >85% and >90% success rates for the non-targeted and targeted attacks, respectively. Owing to the non-targeted UAPs, the DNN models judge most chest X-ray images as COVID-19 cases. The targeted UAPs allow the DNN models to classify most chest X-ray images into a specified target class. The results indicate that careful consideration is required in practical applications of DNNs to COVID-19 diagnosis; in particular, they emphasize the need for strategies to address security concerns. As an example, we show that iterative fine-tuning of DNN models using UAPs improves the robustness of DNN models against UAPs.

Network controllability is an important topic in wide-ranging research fields. However, the relationship between controllability and network structure is poorly understood, although degree heterogeneity is known to determine the controllability. We focus on the size of a minimum dominating set (MDS), a measure of network controllability, and investigate the effect of degree-degree correlation, which is universally observed in real-world networks, on the size of an MDS. We show that disassortativity or negative degree-degree correlation reduces the size of an MDS using analytical treatments and numerical simulation, whereas positive correlations hardly affect the size of an MDS. This result suggests that disassortativity enhances network controllability. Furthermore, apart from the controllability issue, the developed techniques provide new ways of analyzing complex networks with degree-degree correlations.

Theoretical studies have indicated that nestedness and modularity-non-random structural patterns of ecological networks-influence the stability of ecosystems against perturbations; as such, climate change and human activity, as well as other sources of environmental perturbations, affect the nestedness and modularity of ecological networks. However, the effects of climate change and human activities on ecological networks are poorly understood. Here, we used a spatial analysis approach to examine the effects of climate change and human activities on the structural patterns of food webs and mutualistic networks, and found that ecological network structure is globally affected by climate change and human impacts, in addition to current climate. In pollination networks, for instance, nestedness increased and modularity decreased in response to increased human impacts. Modularity in seed-dispersal networks decreased with temperature change (i.e., warming), whereas food web nestedness increased and modularity declined in response to global warming. Although our findings are preliminary owing to data-analysis limitations, they enhance our understanding of the effects of environmental change on ecological communities.

In this study, the metabolic and physiological potential evaluator system based on Kyoto Encyclopedia of Genes and Genomes (KEGG) functional modules was employed to establish a functional classification of archaeal species and to determine the comprehensive functions (functionome) of the previously uncultivated thermophile "Candidatus Caldiarchaeum subterraneum" (Ca. C. subterraneum). A phylogenetic analysis based on the concatenated sequences of proteins common among 142 archaea and 2 bacteria, and among 137 archaea and 13 unicellular eukaryotes suggested that Ca. C. subterraneum is closely related to thaumarchaeotic species. Consistent with the results of the phylogenetic analysis, clustering and principal component analyses based on the completion ratio patterns for all KEGG modules in 79 archaeal species suggested that the overall metabolic and physiological potential of Ca. C. subterraneum is similar to that of thaumarchaeotic species. However, Ca. C. subterraneum possessed almost no genes in the modules required for nitrification and the hydroxypropionate-hydroxybutyrate cycle for carbon fixation, unlike thaumarchaeotic species. However, it possessed all genes in the modules required for central carbohydrate metabolism, such as glycolysis, pyruvate oxidation, the tricarboxylic acid (TCA) cycle, and the glyoxylate cycle, as well as multiple sets of sugar and branched chain amino acid ABC transporters. These metabolic and physiological features appear to support the predominantly aerobic character of Ca. C. subterraneum, which lives in a subsurface thermophilic microbial mat community with a heterotrophic lifestyle.

The hypothesis that variability in natural habitats promotes modular organization is widely accepted for cellular networks. However, results of some data analyses and theoretical studies have begun to cast doubt on the impact of habitat variability on modularity in metabolic networks. Therefore, we re-evaluated this hypothesis using statistical data analysis and current metabolic information. We were unable to conclude that an increase in modularity was the result of habitat variability. Although horizontal gene transfer was also considered because it may contribute for survival in a variety of environments, closely related to habitat variability, and is known to be positively correlated with network modularity, such a positive correlation was not concluded in the latest version of metabolic networks. Furthermore, we demonstrated that the previously observed increase in network modularity due to habitat variability and horizontal gene transfer was probably due to a lack of available data on metabolic reactions. Instead, we determined that modularity in metabolic networks is dependent on species growth conditions. These results may not entirely discount the impact of habitat variability and horizontal gene transfer. Rather, they highlight the need for a more suitable definition of habitat variability and a more careful examination of relationships of the network modularity with horizontal gene transfer, habitats, and environments.

Single amino acid variants (SAVs) are the most abundant form of known genetic variations associated with human disease. Successful prediction of the functional impact of SAVs from sequences can thus lead to an improved understanding of the underlying mechanisms of why a SAV may be associated with certain disease. In this work, we constructed a high-quality structural dataset that contained 679 high-quality protein structures with 2,048 SAVs by collecting the human genetic variant data from multiple resources and dividing them into two categories, i.e., disease-associated and neutral variants. We built a two-stage random forest (RF) model, termed as FunSAV, to predict the functional effect of SAVs by combining sequence, structure and residue-contact network features with other additional features that were not explored in previous studies. Importantly, a two-step feature selection procedure was proposed to select the most important and informative features that contribute to the prediction of disease association of SAVs. In cross-validation experiments on the benchmark dataset, FunSAV achieved a good prediction performance with the area under the curve (AUC) of 0.882, which is competitive with and in some cases better than other existing tools including SIFT, SNAP, Polyphen2, PANTHER, nsSNPAnalyzer and PhD-SNP. The sourcecodes of FunSAV and the datasets can be downloaded at http://sunflower.kuicr.kyoto-u.ac.jp/sjn/FunSAV.

Zinc-binding proteins are the most abundant metalloproteins in the Protein Data Bank where the zinc ions usually have catalytic, regulatory or structural roles critical for the function of the protein. Accurate prediction of zinc-binding sites is not only useful for the inference of protein function but also important for the prediction of 3D structure. Here, we present a new integrative framework that combines multiple sequence and structural properties and graph-theoretic network features, followed by an efficient feature selection to improve prediction of zinc-binding sites. We investigate what information can be retrieved from the sequence, structure and network levels that is relevant to zinc-binding site prediction. We perform a two-step feature selection using random forest to remove redundant features and quantify the relative importance of the retrieved features. Benchmarking on a high-quality structural dataset containing 1,103 protein chains and 484 zinc-binding residues, our method achieved >80% recall at a precision of 75% for the zinc-binding residues Cys, His, Glu and Asp on 5-fold cross-validation tests, which is a 10%-28% higher recall at the 75% equal precision compared to SitePredict and zincfinder at residue level using the same dataset. The independent test also indicates that our method has achieved recall of 0.790 and 0.759 at residue and protein levels, respectively, which is a performance better than the other two methods. Moreover, AUC (the Area Under the Curve) and AURPC (the Area Under the Recall-Precision Curve) by our method are also respectively better than those of the other two methods. Our method can not only be applied to large-scale identification of zinc-binding sites when structural information of the target is available, but also give valuable insights into important features arising from different levels that collectively characterize the zinc-binding sites. The scripts and datasets are available at http://protein.cau.edu.cn/zincidentifier/.

Network modularity is an important structural feature in metabolic networks. A previous study suggested that the variability in natural habitat promotes metabolic network modularity in bacteria. However, since many factors influence the structure of the metabolic network, this phenomenon might be limited and there may be other explanations for the change in metabolic network modularity. Therefore, we focus on archaea because they belong to another domain of prokaryotes and show variability in growth conditions (e.g., trophic requirement and optimal growth temperature), but not in habitats because of their specialized growth conditions (e.g., high growth temperature). The relationship between biological features and metabolic network modularity is examined in detail. We first show the absence of a relationship between network modularity and habitat variability in archaea, as archaeal habitats are more limited than bacterial habitats. Although this finding implies the need for further studies regarding the differences in network modularity, it does not contradict previous work. Further investigations reveal alternative explanations. Specifically, growth conditions, trophic requirement, and optimal growth temperature, in particular, affect metabolic network modularity. We have discussed the mechanisms for the growth condition-dependant changes in network modularity. Our findings suggest different explanations for the changes in network modularity and provide new insights into adaptation and evolution in metabolic networks, despite several limitations of data analysis.