BACKGROUND Heightened surveillance of acute febrile illness in China since 2009 has led to the identification of a severe fever with thrombocytopenia syndrome (SFTS) with an unknown cause. Infection with Anaplasma phagocytophilum has been suggested as a cause, but the pathogen has not been detected in most patients on laboratory testing. METHODS We obtained blood samples from patients with the case definition of SFTS in six provinces in China. The blood samples were used to isolate the causal pathogen by inoculation of cell culture and for detection of viral RNA on polymerase-chain-reaction assay. The pathogen was characterized on electron microscopy and nucleic acid sequencing. We used enzyme-linked immunosorbent assay, indirect immunofluorescence assay, and neutralization testing to analyze the level of virus-specific antibody in patients’ serum samples. RESULTS We isolated a novel virus, designated SFTS bunyavirus, from patients who presented with fever, thrombocytopenia, leukocytopenia, and multiorgan dysfunction. RNA sequence analysis revealed that the virus was a newly identified member of the genus phlebovirus in the Bunyaviridae family. Electron-microscopical examination revealed virions with the morphologic characteristics of a bunyavirus. The presence of the virus was confirmed in 171 patients with SFTS from six provinces by detection of viral RNA, specific antibodies to the virus in blood, or both. Serologic assays showed a virus-specific immune response in all 35 pairs of serum samples collected from patients during the acute and convalescent phases of the illness. CONCLUSIONS A novel phlebovirus was identified in patients with a life-threatening illness associated with fever and thrombocytopenia in China. (Funded by the China Mega-Project for Infectious Diseases and others.)

Abstract. This paper proposes Markovian Generative Adversarial Networks (MGANs), a method for training generative neural networks for efficient texture synthesis. While deep neural network approaches have recently demonstrated remarkable results in terms of synthesis quality, they still come at considerable computational costs (minutes of run-time for low-res images). Our paper addresses this efficiency issue. Instead of a numerical deconvolution in previous work, we precompute a feedforward, strided convolutional network that captures the feature statistics of Markovian patches and is able to directly generate outputs of arbitrary dimensions. Such network can directly decode brown noise to realistic texture, or photos to artistic paintings. With adversarial training, we obtain quality comparable to recent neural texture synthesis methods. As no optimization is required any longer at generation time, our run-time performance (0.25M pixel images at 25Hz) surpasses previous neural texture synthesizers by a significant margin (at least 500 times faster). We apply this idea to texture synthesis, style transfer, and video stylization.

Highlights d Heterogeneity and plasticity of non-parenchymal cells in healthy and NASH liver d Landscape of intrahepatic ligand-receptor signaling at single-cell resolution d Emergence of Trem2+ NASH-associated macrophages (NAMs) in mouse and human NASH d Stellakine secretion and contractile response to vasoactive hormones by HSCs

This paper studies a combination of generative Markov random field (MRF) models and discriminatively trained deep convolutional neural networks (dCNNs) for synthesizing 2D images. The generative MRF acts on higher-levels of a dCNN feature pyramid, controling the image layout at an abstract level. We apply the method to both photographic and non-photo-realistic (artwork) synthesis tasks. The MRF regularizer prevents over-excitation artifacts and reduces implausible feature mixtures common to previous dCNN inversion approaches, permitting synthezing photographic content with increased visual plausibility. Unlike standard MRF-based texture synthesis, the combined system can both match and adapt local features with considerable variability, yielding results far out of reach of classic generative MRF methods.

Implicit methods for modeling protein electrostatics require dielectric properties of the system to be known, in particular, the value of the dielectric constant of protein. While numerous values of the internal protein dielectric constant were reported in the literature, still there is no consensus of what the optimal value is. Perhaps this is due to the fact that the protein dielectric constant is not a “constant” but is a complex function reflecting the properties of the protein’s structure and sequence. Here, we report an implementation of a Gaussian-based approach to deliver the dielectric constant distribution throughout the protein and surrounding water phase by utilizing the 3D structure of the corresponding macromolecule. In contrast to previous reports, we construct a smooth dielectric function throughout the space of the system to be modeled rather than just constructing a “Gaussian surface” or smoothing molecule–water boundary. Analysis on a large set of proteins shows that (a) the average dielectric constant inside the protein is relatively low, about 6–7, and reaches a value of about 20–30 at the protein’s surface, and (b) high average local dielectric constant values are associated with charged residues while low dielectric constant values are automatically assigned to the regions occupied by hydrophobic residues. In terms of energetics, a benchmarking test was carried out against the experimental pKa’s of 89 residues in staphylococcal nuclease (SNase) and showed that it results in a much better RMSD (= 1.77 pK) than the corresponding calculations done with a homogeneous high dielectric constant with an optimal value of 10 (RMSD = 2.43 pK).

The abnormally high number of centrosomes found in many human tumor cells can lead directly to aneuploidy and genomic instability through the formation of multipolar mitotic spindles. To facilitate investigation of the mechanisms that control centrosome reproduction, a frog egg extract arrested in S phase of the cell cycle that supported repeated assembly of daughter centrosomes was developed. Multiple rounds of centrosome reproduction were blocked by selective inactivation of cyclin-dependent kinase 2-cyclin E (Cdk2-E) and were restored by addition of purified Cdk2-E. Confocal immunomicroscopy revealed that cyclin E was localized at the centrosome. These results demonstrate that Cdk2-E activity is required for centrosome duplication during S phase and suggest a mechanism that could coordinate centrosome reproduction with cycles of DNA synthesis and mitosis.

Figure 1. HoloGAN learns to separate pose from identity (shape and appearance) only from unlabelled 2D images without sacrificing the visual fidelity of the generated images. All results shown here are sampled from HoloGAN for the same identities in each row but in different poses. AbstractWe propose a novel generative adversarial network (GAN) for the task of unsupervised learning of 3D representations from natural images. Most generative models rely on 2D kernels to generate images and make few assumptions about the 3D world. These models therefore tend to create blurry images or artefacts in tasks that require a strong 3D understanding, such as novel-view synthesis. HoloGAN instead learns a 3D representation of the world, and to render this representation in a realistic manner. Unlike other GANs, HoloGAN provides explicit control over the pose of generated objects through rigid-body transformations of the learnt 3D features. Our experiments show that using explicit 3D features enables HoloGAN to disentangle 3D pose and identity, which is further decomposed into shape and appearance, while still being able to generate images with similar or higher visual quality than other generative models. HoloGAN can be trained end-to-end from unlabelled 2D images only. In particular, we do not require pose labels, 3D shapes, or multiple views of the same objects. This shows that HoloGAN is the first generative model that learns 3D representations from natural images in an entirely unsupervised manner.

BackgroundAnlotinib is a novel multi-target tyrosine kinase inhibitor that is designed to primarily inhibit VEGFR2/3, FGFR1-4, PDGFR α/β, c-Kit, and Ret. We aimed to evaluate the safety, pharmacokinetics, and antitumor activity of anlotinib in patients with advanced refractory solid tumors.MethodsAnlotinib (5–16 mg) was orally administered in patients with solid tumor once a day on two schedules: (1) four consecutive weeks (4/0) or (2) 2-week on/1-week off (2/1). Pharmacokinetic sampling was performed in all patients. Twenty-one patients were further enrolled in an expanded cohort study on the recommended dose and schedule. Preliminary tumor response was also assessed.ResultsOn the 4/0 schedule, dose-limiting toxicity (DLT) was grade 3 hypertension at 10 mg. On the 2/1 schedule, DLT was grade 3 hypertension and grade 3 fatigue at 16 mg. Pharmacokinetic assessment indicated that anlotinib had long elimination half-lives and significant accumulation during multiple oral doses. The 2/1 schedule was selected, with 12 mg once daily as the maximum tolerated dose for the expanding study. Twenty of the 21 patients (with colon adenocarcinoma, non-small cell lung cancer, renal clear cell cancer, medullary thyroid carcinoma, and soft tissue sarcoma) were assessable for antitumor activity of anlotinib: 3 patients had partial response, 14 patients had stable disease including 12 tumor burden shrinkage, and 3 had disease progression. The main serious adverse effects were hypertension, triglyceride elevation, hand-foot skin reaction, and lipase elevation.ConclusionsAt the dose of 12 mg once daily at the 2/1 schedule, anlotinib displayed manageable toxicity, long circulation, and broad-spectrum antitumor potential, justifying the conduct of further studies.Electronic supplementary materialThe online version of this article (doi:10.1186/s13045-016-0332-8) contains supplementary material, which is available to authorized users.