Search

Embodied Question Answering (EQA) is a newly defined research area where an agent is required to answer the users questions by exploring the real-world environment. It has attracted increasing research interests due to its broad applications in personal assistants and in-home robots. Most of the existing methods perform poorly in terms of answering and navigation accuracy due to the absence of fine-level semantic information, stability to the ambiguity, and 3D spatial information of the virtual environment. To tackle these problems, we propose a depth and segmentation based visual attention mechanism for Embodied Question Answering. Firstly, we extract local semantic features by introducing a novel high-speed video segmentation framework. Then guided by the extracted semantic features, a depth and segmentation based visual attention mechanism is proposed for the Visual Question Answering (VQA) sub-task. Further, a feature fusion strategy is designed to guide the navigators training process without much additional computational cost. The ablation experiments show that our method effectively boosts the performance of the VQA module and navigation module, leading to 4.9% and 5.6% overall improvement in EQA accuracy on House3D and Matterport3D datasets respectively.

Albumin-based hydrogels have emerged as promising nanoparticle systems for the effective delivery of hydrophobic anticancer drugs. Anti-cancer drugs often cause some adverse effects, such as toxicity and rapid clearance by mononuclear phagocytic systems. Herein, a new strategy of synthesizing N-hydroxysuccinimide (NHS)-activated linker to form crosslinkable albumin-based hydrogels (CABH) is reported. The CABH favored physiochemical characteristics improvement of doxorubicin (Dox) and drug release. The CABH was constructed depending on the crosslinking reaction between NHS activated glycerol and albumin. The size of CABH was approximately 200 nm examined by dynamic light scattering (DLS) and transmission electron microscopy (TEM). It was found that the particle size and size distribution of the CABH remained stable in neutral PBS for 1 week. Dox loaded CABH would be controllably released in weak acidic environment verified by in vitro release and in vitro cell imaging. The Dox loaded hydrogel results in significant killing in the case of acidic culture medium. Our work provides a crosslinking method to formulate albumin nanoplatform and improve the size, stability, drug loading capacity and controlled release, which throws light on the potential application in drug delivery.

Background N6-methyladenosine (m6A) is the most common modification on RNAs and LncRNAs. It plays an important role in cancer stem cell differentiation, T cell differentiation, and immune homeostasis. In this study, we explored the potential roles of m6A modification of RNA in melanoma and investigated the immune cell infiltration in tumor microenvironment in diverse m6Aclusters and different m6Ascore groups. Methods The gene expression data and corresponding clinical information of melanoma patients were downloaded from TCGA and GEO databases (GSE65904). A consensus clustering algorithm determined m6A modification patterns based on 14 m6A regulators. Gene set variation analysis (GSVA) explored the biological process and single sample gene set enrichment analyses (ssGSEA) quantified the relative abundance of each immune cell infiltration. m6A-related differentially expressed genes (DEGs) were identified based on three patterns found among patients. An m6A-related gene signature (m6Ascore) was constructed based on genes with significant prognostic value using the principal component analysis algorithm. Results Three m6A modification patterns were identified based on 14 m6A regulators, named as m6Aclusters A-C. The prognosis of m6Acluster A was more favorable than m6Aclusters B and C, and it was more closely associated with immune regulation. After screening 4430 DEGs, three m6A modification genomic phenotypes were identified based on 845 prognostic DEGs. To quantify the m6A modification patterns of individual tumor, an m6Ascore was constructed, and patients were classified into high and low m6Ascore groups. The low m6Ascore group, which had a favorable prognosis, was more relevant to immunology. The expression of PD-L1 was higher and the immunophenoscore (IPS) revealed stronger response to immunotherapy in the low m6Ascore group. Conclusions This study identified three m6A modification patterns with different immune characteristics in melanoma patients and constructed an m6Ascore system to predict prognosis and immunogenicity of patients, which is conducive to clinical prognosis judgment and individual treatment.

In an intelligent transportation system, the key problem of traffic forecasting is how to extract periodic temporal dependencies and complex spatial correlations. Current state-of-the-art methods for predicting traffic flow are based on graph architectures and sequence learning models, but they do not fully exploit dynamic spatial-temporal information in the traffic system. Specifically, the temporal dependencies in the short-range are diluted by recurrent neural networks. Moreover, local spatial information is also ignored by existing sequence models, because their convolution operation uses global average pooling. Besides, accidents may occur during object transition, which will cause congestion in the real world and further decrease prediction accuracy. To overcome these challenges, we propose Spatial-Temporal Conv-sequence Learning (STCL), where a focused temporal block uses unidirectional convolution to capture short-term periodic temporal dependencies effectively, and a patial-temporal fusion module is responsible for extracting dependencies of interactions and decreasing the feature dimensions. Moreover, as the accidents features have an impact on local traffic congestion, we employ position encoding to detect anomalies in complex traffic situations. We have conducted a large number of experiments on real-world tasks and verified the effectiveness of our proposed method.

E3 ubiquitin ligases (E3s), the second most common cancer-related functional protein family, play vital roles in multiple tumors. However, their importance in prognosis and immunotherapy of lung adenocarcinoma (LUAD) is not clear. First, utilizing the data from The Cancer Genome Atlas (TCGA), we comprehensively assessed the expression profile and immunological association of 13 E3s in LUAD patients. Consequently, MARCH1 was considered a candidate for further study. Second, several algorithms were applied to assess the correlation between MARCH1 and immunological characteristics in the LUAD tumor microenvironment. Third, an immune risk score (IRS) was developed to predict the prognosis. Finally, the immunological relationship of MARCH1 in pan-cancer was also estimated. We found that E3s were disordered in LUAD. Among them, MARCH1 was positively correlated with most immunological characteristics, indicating that MARCH1 designed an inflamed TME in LUAD. Coincidently, LUAD with low MARCH1 expression had a poor prognosis and was not sensitive to immune checkpoint blockers. In addition, the IRS could accurately predict the prognosis. In pan-cancer, MARCH1 was also positively correlated with most immunological characteristics. In conclusion, MARCH1 could be a novel and promising biomarker for immune status and effectiveness of immunotherapy for LUAD patients.

Background Helicobacter pylori (H. pylori) infection is closely associated with the tumorigenesis of gastric cancer. The aim of the present study was to identify the key regulator in H. pylori-related gastric cancer and to study the expression level and clinical value of the indicated key regulator in gastric cancer. Methods The GSE6143 dataset was used to identify differentially expressed genes (DEGs) with limma R package, and enrichment analysis was done using the Metascape web-based portal. The protein-protein interaction analysis was done using Search Tool for the Retrieval of Interacting Genes/Proteins. Gastric adenocarcinoma AGS and BGC-823 cells were treated with H. pylori strain 26695 to construct the in vitro H. pylori infection model, and quantitative reverse transcription polymerase chain reaction was used to analyze the mRNA levels of indicated genes. The correlation analysis between two genes in gastric cancer was done by GEPIA. Furthermore, the PTPRC expression by pathological features analysis was conducted in UALCAN, an easy to use, interactive web-portal (http://ualcan.path.uab.edu). The survival analysis for gastric cancer, based on PTPRC expression levels, was done using the Kaplan-Meier plotter. Results DEGs in gastric mucosa with or without H. pylori infection were identified and enriched in immune-related pathways and cancer pathways. The protein-protein interaction analysis confirmed the enrichment analysis of gene ontology. H. pylori strain 26695 exposure also confirmed the alteration of gene expression levels in AGS and BGC-823 cells. PTPRC was co-expressed with CSF2RB and TNFRSF7, indicating a significant positive correlation in gastric cancer. PTPRC was overexpressed in gastric cancer, and the overexpression of PTPRC was positively correlated with the progression of gastric cancer. Furthermore, the high expression of PTPRC could act as a poor prognostic factor for gastric cancer patients, especially for those at advanced stage. Conclusions H. pylori-induced PTPRC is overexpressed in gastric cancer, and the overexpression of PTPRC is positively associated with the development of gastric cancer. The high expression of PTPRC could serve as poor prognostic biomarker for gastric cancer patients, especially for those at advanced stage. H. pylori-induced PTPRC is a prognostic biomarker for gastric cancer.

Due to the effects of climate changing, the importance of outdoor thermal comfort has been recognized, and has gained more and more research attentions. Unlike indoor space where air conditioning can be easily implemented, outdoor thermal comfort can only be achieved by localized thermal management. Using textile is a simple but energy-saving way to realize outdoor thermal comfort. Herein, we report the design of a smart thermal management film with the silver/vanadium dioxide/silver (Ag/VO2/Ag) sandwich structure prepared by one-dimensional (1 D) nanowires. It was found that the Ag/VO2/Ag sandwich film was able to lower the temperature by around 10 °C under intense infrared (IR) radiation. In addition, the Ag/VO2/Ag sandwich structure film showed a thermo-responsive electrical conductivity and an outstanding bending stability, due to network structure formed by nanowires. It was experimentally proved that this sandwich structure was superior to other layer structures in IR shielding performance and thermo-responsive electrical conductivity. The as-prepared Ag/VO2/Ag sandwich structure film has great potential for various applications such as wearable devices, flexible electronics, medical monitors and smart IR radiation management.

The type-I interferon (IFN-I) signaling pathway plays pivot roles in defending against pathogen invasion. Exogenous ssRNA and dsRNA could be immunogenic. RNA-mediated IFN signaling is extensively studied in the field. The incorrect functioning of this pathway leads to either autoimmune diseases or suffering from microorganism invasion. From the discrimination of “self” and “non-self” molecules by receptors to the fine-tune modulations in downstream cascades, all steps are under the surveillance featured by complex feedbacks and regulators. Studies in recent years highlighted the emerging roles of long noncoding RNAs (lncRNAs) as a reservoir for signaling regulation. LncRNAs bind to targets through the structure and sequence, and thus the mechanisms of action can be complex and specific. Here, we summarized lncRNAs modulating the RNA-activated IFN-I signaling pathway according to the event order during the signaling. We hope this review help understand how lncRNAs are participating in the regulation of IFN-I signaling.

Taxi arrival time prediction is essential for building intelligent transportation systems. Traditional prediction methods mainly rely on extracting features from traffic maps, which cannot model complex situations and nonlinear spatial and temporal relationships. Therefore, we propose Multi-View Spatial-Temporal Model (MVSTM) to capture the mutual dependence of spatial-temporal relations and trajectory features. Specifically, we use graph2vec to model the spatial view, dual-channel temporal module to model the trajectory view, and structural embedding to model traffic semantics. Experiments on large-scale taxi trajectory data have shown that our approach is more effective than the existing novel methods. The source code can be found at https://github.com/775269512/SIGSPATIAL-2021-GISCUP-4th-Solution., 4 pages, 2 figures

The scene text interpretation is a critical part of the natural scene interpretation. Currently, most of the existing work is based on high-end graphics processing units (GPUs) implementation, which is commonly used on the server side. However, in Internet of Things (IoT) application scenarios, the communication overhead from the edge device to the server is quite large, which sometimes even dominates the total processing time. Hence, the edge-computing oriented design is needed to solve this problem. In this paper, we present an architectural design and implementation of a natural scene text interpretation (NSTI) accelerator, which can classify and localize the text region on pixel-level efficiently in real-time on mobile devices. To target the real-time and low-latency processing, the binary convolutional encoder-decoder network is adopted as the core architecture to enable massive parallelism due to its binary feature. Massively parallelized computations and a highly pipelined data flow control enhance its latency and throughput performance. In addition, all the binarized intermediate results and parameters are stored on chip to eliminate the power consumption and latency overhead of the off-chip communication. The NSTI accelerator is implemented in a 40 nm CMOS technology, which can process scene text images (size of 128 × 32) at 34 fps and latency of 40 ms for pixelwise interpretation with the pixelwise classification accuracy over 90% on ICDAR-03 and ICDAR-13 dataset. The real energy-efficiency is 698 GOP/s/W and the peak energy-efficiency can get up to 7825 GOP/s/W. The proposed accelerator is 7 times more energy efficient than its optimized GPU-based implementation counterpart, while maintaining a real-time throughput with latency of 40 ms. Ministry of Education (MOE) Arizona State University’s work was supported by National Science Foundation under Grant IIS/CPS-1652038. Nanyang Technological Unversity’s work was supported by MOE AcRF Tier 2 under Grant MOE2015-T2-2-013.

Photochromic materials have been extensively studied because they are quite attractive and promising for many applications. Tungsten oxide (WO3), a typical photochromic material, has attracted considerable attention. So far, many advances have been achieved in different subfields of WO3 photochromism, mainly focused on the enhancement of WO3 photochromic properties, which depends not only on the chemical features but also on the microstructures, interface and hybrid components. This review aims to summarize recent progress in different advanced strategies to improve WO3 photochromic properties, such as a fast photoresponse (within 1 min), strong and stable photochromism, visible-light coloration and reversible photochromism. However, it is still challenging to achieve these properties for large-scale commercial applications. In addition, many previous studies have been limited to lab research purposes, although some photochromic WO3-based devices have been fabricated. To address the above-mentioned issues clearly, this review will provide a discussion of the following aspects. First, the fundamental photochromism of WO3 is described. Secondly, various advances towards improving its photochromic properties will be summarized. Thirdly, the optical applications of WO3-based photochromic materials will be reviewed and discussed. And finally, a perspective on the development trends in WO3-based photochromic materials and devices is proposed.

There is great attention to develop hardware accelerator with better energy efficiency, as well as throughput, than GPUs for convolutional neural network (CNN). The existing solutions have relatively limited parallelism as well as large power consumption (including leakage power). In this paper, we present a resistive random access memory (ReRAM)-accelerated CNN that can achieve significantly higher throughput and energy efficiency when the CNN is trained with binary constraints on both weights and activations, and is further mapped on a digital ReRAM-crossbar. We propose an optimized accelerator architecture tailored for bitwise convolution that features massive parallelism with high energy efficiency. Numerical experiment results show that the binary CNN accelerator on a digital ReRAM-crossbar achieves a peak throughput of 792 GOPS at the power consumption of 4.5 mW, which is 1.61 times faster and 296 times more energy-efficient than a high-end GPU. NRF (Natl Research Foundation, S’pore) MOE (Min. of Education, S’pore) Accepted version

Vanadium oxides (VOx) and their related compounds have been received extensive attentions because of their special chemical and physical properties, which make them have a wide range of practical applications. VOx can offer high pseudocapacitance for energy storage due to its high electrochemical activity. Herein, 0D (nanoparticles), 1D (nanowires), 2D (nanosheets) & 3D (nanodonuts) VOx were successfully synthesized with assistance of glucose by a facile one-step hydrothermal method. The relationship between the structure and the electrochemical properties of VOx was investigated in detail. Among all the morphologies, the donut-like VOx displays the best pseudocapacitive performance under our reported conditions, with a specific capacitance as high as 225.8 F g−1 at a scan rate of 2 mV s−1. The synergy effect between multiple vanadium oxidation states (mixed-valence) and the donut-like geometry contributes to electrochemical properties. That is, the mixed-valence offers theoretical specific capacitance and the unique donut-like structure provides a more active surface and supported high-speed ionic and electronic transmission channels. The good electrochemical performance of the 0-3D VOx nanostructures indicates their high potential as electrode materials for supercapacitors.

The recently emerging resistive random-access memory (RRAM) can provide nonvolatile memory storage but also intrinsic computing for matrix-vector multiplication, which is ideal for the low-power and high-throughput data analytics accelerator performed in memory. However, the existing RRAM crossbar--based computing is mainly assumed as a multilevel analog computing, whose result is sensitive to process nonuniformity as well as additional overhead from AD-conversion and I/O. In this article, we explore the matrix-vector multiplication accelerator on a binary RRAM crossbar with adaptive 1-bit-comparator--based parallel conversion. Moreover, a distributed in-memory computing architecture is also developed with the according control protocol. Both memory array and logic accelerator are implemented on the binary RRAM crossbar, where the logic-memory pair can be distributed with the control bus protocol. Experimental results have shown that compared to the analog RRAM crossbar, the proposed binary RRAM crossbar can achieve significant area savings with better calculation accuracy. Moreover, significant speedup can be achieved for matrix-vector multiplication in neural network--based machine learning such that the overall training and testing time can be both reduced. In addition, large energy savings can be also achieved when compared to the traditional CMOS-based out-of-memory computing architecture. NRF (Natl Research Foundation, S’pore) MOE (Min. of Education, S’pore) Accepted version

It is challenging to detect curve texts due to their irregular shapes and varying sizes. In this paper, we first investigate the deficiency of the existing curve detection methods and then propose a novel Conditional Spatial Expansion (CSE) mechanism to improve the performance of curve text detection. Instead of regarding the curve text detection as a polygon regression or a segmentation problem, we treat it as a region expansion process. Our CSE starts with a seed arbitrarily initialized within a text region and progressively merges neighborhood regions based on the extracted local features by a CNN and contextual information of merged regions. The CSE is highly parameterized and can be seamlessly integrated into existing object detection frameworks. Enhanced by the data-dependent CSE mechanism, our curve text detection system provides robust instance-level text region extraction with minimal post-processing. The analysis experiment shows that our CSE can handle texts with various shapes, sizes, and orientations, and can effectively suppress the false-positives coming from text-like textures or unexpected texts included in the same RoI. Compared with the existing curve text detection algorithms, our method is more robust and enjoys a simpler processing flow. It also creates a new state-of-art performance on curve text benchmarks with F-score of up to 78.4$\%$., Comment: This paper has been accepted by IEEE International Conference on Computer Vision and Pattern Recognition (CVPR 2019)

The main research content of this project is the water recovery efficiency of the hot-pressure-driven air water-taking device when the phase-change material is used for energy supply, and combined with the research process, an automatic water-taking device using desert temperature difference and phase-change material storage is designed. At a time when energy sustainability is of great concern, we must make the development of green energy a core goal. Among them, solar energy as a green non-polluting new energy, its development and utilization is an important research topic and direction. The desert areas have plenty of solar energy and enough temperature differences to take advantage of these two resources and combine the lack of fresh water in the area. The purpose of this project is to design a device that uses phase-change materials to store heat to provide heat pressure, guide air directional flow, and use high-efficiency coaxial water collection sails for air water collection, with less human and material input, to obtain the fresh water necessary for life. And from the theoretical and physical level to demonstrate its feasibility, explore the direction of marketing.

With the gradual aggravation of desertification, aquatic plants are very rare, rain is scarce, and the air is very dry in desert areas, which poses unprecedented challenges to the survival of local residents. In order to solve the crisis of lack of fresh water, the research on air water intake was carried out, and the corresponding design of air purification device was carried out. The benefit and feasibility of the air purification device were analyzed by using the model and software.

Background Dysfunction in fibroblast growth factor receptor (FGFR) signaling has been reported in non-small cell lung cancer (NSCLC), particularly in lung squamous cell carcinoma (LUSC). We hereby assessed the frequency of FGFR aberrations in a large Chinese NSCLC population. Methods A total of 10,966 NSCLCs whose tumor specimen and/or circulating cell-free DNA (cfDNA) underwent hybridization capture-based next-generation sequencing (NGS) of 400+ cancer-related genes were retrospectively reviewed. Patients' clinical characteristics and treatment histories were further evaluated. Results FGFR aberrations, including mutations, fusions, and gene amplifications, were detected in approximately 1.9% (210/10,966) of the population. FGFR abnormalities were more frequently observed in LUSCs (6.8%, 65/954) than lung adenocarcinomas (1.3%, 128/9596). Previously reported FGFR oncogenic mutations were identified in 19 patients (~0.17%), of which, 68% were male LUSC patients. Eleven out of the 19 patients (58%) had concurrent altered PI3K signaling, thus, highlighting an intriguing molecular characteristic and potential combination therapeutic strategy of dual-targeting FGFR and PI3K signaling in those patients. Furthermore, FGFR fusions retaining the intact kinase domain were identified in 12 patients (0.11%), including nine FGFR3-TACC3 cases, one FGFR2-INA fusion, one novel FGFR4-RAPGEFL1 rearrangement event, and one novel gene fusion between FGFR1 and the SLC20A2 5'-untranslated region, which may result in FGFR1 overexpression. Concomitant EGFR mutations or EGFR amplifications were observed in six patients, four of whom were previously treated with EGFR tyrosine kinase inhibitors, suggesting that FGFR fusions may act as acquired resistance to anti-EGFR therapies. In addition, FGFR amplification was detected in 24 patients, with FGFR1 amplification being the most common event. Notably, FGFR oncogenic mutations, fusions, and gene amplifications were almost always mutually exclusive events. Conclusion We herein report an incidence of FGFR anomalies in a large Chinese NSCLC population, including mutations, gene amplifications, and previously characterized and novel FGFR fusions involving TACC3, INA, RAPGEFL1, and SLC20A2. Keywords: FGFR, activating mutation, fusion, gene amplification, NSCLC, FGFR inhibitors Citation Format: Zhen Zhou, Zichuan Liu, Qiuxiang Ou, Xue Wu, Xiaonan Wang, Yang Shao, Hongyan Liu, Yu Yang. Landscape of FGFRactivating aberrations in Chinese non-small cell lung cancer [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 745.

Clinical effects of high frequency hyperthermia-assisted irinotecan chemotherapy on patients with middle and advanced colorectal cancer and its safety assessment were investigated. A retrospective analysis was performed on the medical records of 103 patients with middle and advanced colorectal cancer treated in Cancer Center of Guangzhou Medical University from May 2011 to June 2015, including 48 patients receiving irinotecan plus conventional treatment (irinotecan group) and 55 patients receiving high frequency hyperthermia-assisted irinotecan plus conventional treatment (combination group). The treatment effects, severity and incidence of adverse reactions, quality of life and 3-year survival rates of patients were analyzed and compared between the two groups. After 4 courses of treatment, there were statistically significant differences in the proportions of patients with partial remission and objective remission, with those in combination group higher than those in irinotecan group (both P

A new approach for real-time scene text recognition is proposed in this paper. A novel binary convolutional encoder-decoder network (B-CEDNet) together with a bidirectional recurrent neural network (Bi-RNN). The B-CEDNet is engaged as a visual front-end to provide elaborated character detection, and a back-end Bi-RNN performs character-level sequential correction and classification based on learned contextual knowledge. The front-end B-CEDNet can process multiple regions containing characters using a one-off forward operation, and is trained under binary constraints with significant compression. Hence it leads to both remarkable inference run-time speedup as well as memory usage reduction. With the elaborated character detection, the back-end Bi-RNN merely processes a low dimension feature sequence with category and spatial information of extracted characters for sequence correction and classification. By training with over 1,000,000 synthetic scene text images, the B-CEDNet achieves a recall rate of 0.86, precision of 0.88 and F-score of 0.87 on ICDAR-03 and ICDAR-13. With the correction and classification by Bi-RNN, the proposed real-time scene text recognition achieves state-of-the-art accuracy while only consumes less than 1-ms inference run-time. The flow processing flow is realized on GPU with a small network size of 1.01 MB for B-CEDNet and 3.23 MB for Bi-RNN, which is much faster and smaller than the existing solutions.

Convolutional neural network (CNN) based machine learning requires a highly parallel as well as low power consumption (including leakage power) hardware accelerator. In this paper, we will present a digital ReRAM crossbar based CNN accelerator that can achieve significantly higher throughput and lower power consumption than state-of-arts. The CNN is trained with binary constraints on both weights and activations such that all operations become bitwise. With further use of 1-bit comparator, the bitwise CNN model can be naturally realized on a digital ReRAM-crossbar device. A novel sneak-path-free ReRAM-crossbar is further utilized for large-scale realization. Simulation experiments show that the bitwise CNN accelerator on the digital ReRAM crossbar achieves 98.3% and 91.4% accuracy on MNIST and CIFAR-10 benchmarks, respectively. Moreover, it has a peak throughput of 792GOPS at the power consumption of 6.3mW, which is 18.86 times higher throughput and 44.1 times lower power than CMOS CNN (non-binary) accelerators.

Generic and scalable data analysis procedures are highly demanded by the increasing number of multi-dimensional biomedical data. However, especially for time-lapse biological data, the high level of noise prevents for automated high-throughput analysis methods. The rapid developing of machine-learning methods and particularly deep-learning methods provide new tools and methodologies that can help in the denoising of such data. Using a convolutional encoder-decoder network, one can provide a scalable bio-image platform, called NucleiNet, to automatically segment, classify and track cell nuclei. The proposed method can achieve 0.99 F-score and 0.99 pixel-wise accuracy on C. elegans dataset, which means that over 99% of nuclei can be successfully detected with no merging nuclei found.

FPGA-based hardware accelerator for convolutional neural networks (CNNs) has obtained great attentions due to its higher energy efficiency than GPUs. However, it has been a challenge for FPGA-based solutions to achieve a higher throughput than GPU counterparts. In this paper, we demonstrate that FPGA acceleration can be a superior solution in terms of both throughput and energy efficiency when a CNN is trained with binary constraints on weights and activations. Specifically, we propose an optimized accelerator architecture tailored for bitwise convolution and normalization that features massive spatial parallelism with deep pipeline (temporal parallelism) stages. Experiment results show that the proposed architecture running at 90 MHz on a Xilinx Virtex-7 FPGA achieves a computing throughput of 7.663 TOPS with a power consumption of 8.2 W regardless of the batch size of input data. This is 8.3x faster and 75x more energy-efficient than a Titan X GPU for processing online individual requests (in small batch size). For processing static data (in large batch size), the proposed solution is on a par with a Titan X GPU in terms of throughput while delivering 9.5x higher energy efficiency.

FPGA-based hardware accelerators for convolutional neural networks (CNNs) have received attention due to their higher energy efficiency than GPUs. However, it is challenging for FPGA-based solutions to achieve a higher throughput than GPU counterparts. In this article, we demonstrate that FPGA acceleration can be a superior solution in terms of both throughput and energy efficiency when a CNN is trained with binary constraints on weights and activations. Specifically, we propose an optimized fully mapped FPGA accelerator architecture tailored for bitwise convolution and normalization that features massive spatial parallelism with deep pipelines stages. A key advantage of the FPGA accelerator is that its performance is insensitive to data batch size, while the performance of GPU acceleration varies largely depending on the batch size of the data. Experiment results show that the proposed accelerator architecture for binary CNNs running on a Virtex-7 FPGA is 8.3× faster and 75× more energy-efficient than a Titan X GPU for processing online individual requests in small batch sizes. For processing static data in large batch sizes, the proposed solution is on a par with a Titan X GPU in terms of throughput while delivering 9.5× higher energy efficiency.

Exploring theory and methods to analyze the impacts of non real-time services on jitter performance of real-time services is a quite challenging but meaningful job. This article puts forward a general theoretical method to calculate packet jitter in network node bearing mixed services. Based on queuing theory and Markov theory, the network node is modeled as a double-queue single-server and limited-cache queuing system with thresholds. Two-state Markov-modulated Bernoulli process (MMBP-2) and interrupted Bernoulli process (IBP) are used to model the arrival processes of real-time services and small data services respectively in a packet switched network. In order to depict the interaction between real-time services and small data services, a four-dimensional discrete-time Markov chain is implemented to describe the transition of the system states. By solving the system model, expressions for packet jitter are obtained. The results given by the model are then compared with the simulation results obtained by network simulator, version 3 (NS-3).

Correction for ‘Advances on tungsten oxide based photochromic materials: strategies to improve their photochromic properties’ by Shufen Wang et al., J. Mater. Chem. C, 2018, 6, 191–212.

In mammals, sex differentiation of primordial germ cells (PGCs) is determined by extrinsic cues from the environment. In mouse female PGCs, expression of stimulated by retinoic acid gene 8 (Stra8) and meiosis are induced in response to retinoic acid provided from the mesonephroi. Given the widespread role of retinoic acid signalling during development, the molecular mechanisms that enable PGCs to express Stra8 and enter meiosis in a timely manner are unknown. Here we identify gene-dosage-dependent roles in PGC development for Ring1 and Rnf2, two central components of the Polycomb repressive complex 1 (PRC1). Both paralogues are essential for PGC development between days 10.5 and 11.5 of gestation. Rnf2 is subsequently required in female PGCs to maintain high levels of Oct4 (also known as Pou5f1) and Nanog expression, and to prevent premature induction of meiotic gene expression and entry into meiotic prophase. Chemical inhibition of retinoic acid signalling partially suppresses precocious Oct4 downregulation and Stra8 activation in Rnf2-deficient female PGCs. Chromatin immunoprecipitation analyses show that Stra8 is a direct target of PRC1 and PRC2 in PGCs. These data demonstrate the importance of PRC1 gene dosage in PGC development and in coordinating the timing of sex differentiation of female PGCs by antagonizing extrinsic retinoic acid signalling.

Future data-oriented computing requires intensive memory access with burden of I/O interconnection. Traditional wired interconnect has limited bandwidth and energy-efficiency with no feature of configurability. This paper introduces an adaptive sub-THz wireless interconnect between cores and main memory (DRAMs) using MIMO beamforming. To satisfy the requirement of energy efficiency and latency, a configuration problem for wireless channel management at memory controller is formulated to find the optimal value of the beamforming matrix. A low latency beamforming algorithm for sub-THz based interconnect is proposed. To meet the requirement of optimal power efficiency and nanosecond level beamforming latency, a gradient-free Mesh Adaptive Direct Search (MADS) based method is designed to find the sub-optimal solution for beamforming without knowing the channel information. The simulation shows that the proposed MADS-based algorithm converges rapidly and achieves average 90.3% optimal power efficiency when compared to the solution provided by gradient-based convex optimization algorithm. Moreover, we shows that wireless interconnect can provide better energy efficiency compared with the PCB trace wired counterparts under the same data rate.

Oocytes develop the competence for meiosis and early embryogenesis during their growth. Setdb1 is a histone H3 lysine 9 (H3K9) methyltransferase required for post-implantation development and has been implicated in the transcriptional silencing of genes and endogenous retroviral elements (ERVs). To address its role in oogenesis and pre-implantation development, we conditionally deleted Setdb1 in growing oocytes. Loss of Setdb1 expression greatly impaired meiosis. It delayed meiotic resumption, altered the dynamics of chromatin condensation, and impaired kinetochore-spindle interactions, bipolar spindle organization, and chromosome segregation in more mature oocytes. The observed phenotypes related to changes in abundance of specific transcripts in mutant oocytes. Setdb1 maternally deficient embryos arrested during pre-implantation development and showed comparable defects during cell cycle progression and in chromosome segregation. Finally, transcriptional profiling data indicate that Setdb1 down-regulates rather than silences expression of ERVK and ERVL-MaLR retrotransposons and associated chimearic transcripts during oogenesis. Our results identify Setdb1 as a novel meiotic and embryonic competence factor safeguarding genome integrity at the onset of life.

International audience; Somatic cell nuclear transfer (SCNT) has been successfully used in many species to produce live cloned offspring, albeit with low efficiency. The low frequency of successful development has usually been ascribed to incomplete or inappropriate reprogramming of the transferred nuclear genome. Elucidating the genetic differences between normal fertilized and cloned embryos is key to understand the low efficiency of SCNT. Here, we show that expression of HSPC117, which encodes a hypothetical protein of unknown function, was absent or very low in cloned mouse blastocysts. To investigate the role of HSPC117 in embryo development, we knocked-down this gene in normal fertilized embryos using RNA interference. We assessed the post-implantation survival of HSPC117 knock-down embryos at 3 stages: E9 (prior to placenta formation); E12 (after the placenta was fully functional) and E19 (post-natal). Our results show that, although siRNA-treated in vivo fertilized/produced (IVP) embryos could develop to the blastocyst stage and implanted without any difference from control embryos, the knock-down embryos showed substantial fetal death, accompanied by placental blood clotting, at E12. Furthermore, comparison of HSPC117 expression in placentas of nuclear transfer (NT), intracytoplasmic sperm injection (ICSI) and IVP embryos confirmed that HSPC117 deficiency correlates well with failures in embryo development: all NT embryos with a fetus, as well as IVP and ICSI embryos, had normal placental HSPC117 expression while those NT embryos showing reduced or no expression of HSPC117 failed to form a fetus. In conclusion, we show that HSPC117 is an important gene for post-implantation development of embryos, and that HSPC117 deficiency leads to fetal abnormalities after implantation, especially following placental formation. We suggest that defects in HSPC117 expression may be an important contributing factor to loss of cloned NT embryos in vivo.

The mature oocyte contains the full complement of maternal proteins required for fertilization, the transition to zygotic transcription, and the beginning stages of embryogenesis. Many of these proteins have yet to be characterized. In this study, two-dimensional electrophoresis (2-DE) of mouse metaphase-II (MII) oocyte proteins, stained with silver staining or Pro-Q Diamond dye, was performed to describe the proteome and phosphoproteome of the mouse oocyte derived from ICR mice. A total of 869 selected protein spots, corresponding to 380 unique proteins, were identified successfully by mass spectrometry, in which 90 protein spots representing 53 unique proteins have been stained with Pro-Q Diamond, indicating that they are in phosphorylated forms. All identified proteins were bioinformatically annotated in detail and compared with the embryonic stem cell (ESC) proteome. A proteome reference database for the mouse oocyte was established from the protein data generated in this study, which can be accessed over the Internet ( http://reprod.njmu.edu.cn/2d). This database is the most detailed mouse oocyte proteomic database to date. It should be valuable in expanding our knowledge of the regulation of signaling in oogenesis, fertilization, and embryo development, while revealing potential mechanisms for epigenetic reprogramming.

Therapeutic cloning, whereby embryonic stem cells (ESCs) are derived from nuclear transfer (NT) embryos, may play a major role in the new era of regenerative medicine. In this study we established forty nuclear transfer-ESC (NT-ESC) lines that were derived from NT embryos of different donor cell types or passages. We found that NT-ESCs were capable of forming embryoid bodies. In addition, NT-ESCs expressed pluripotency stem cell markers in vitro and could differentiate into embryonic tissues in vivo. NT embryos from early passage R1 donor cells were able to form full term developed pups, whereas those from late passage R1 ES donor cells lost the potential for reprogramming that is essential for live birth. We subsequently established sequential NT-R1-ESC lines that were developed from NT blastocyst of late passage R1 ESC donors. However, these NT-R1-ESC lines, when used as nuclear transfer donors at their early passages, failed to result in live pups. This indicates that the therapeutic cloning process using sequential NT-ESCs may not rescue the developmental deficiencies that resided in previous donor generations.

During mouse preimplantation development, major changes in cell fate are accompanied by extensive alterations of gene expression programs. Embryos first transition from a maternal to zygotic program and subsequently specify the pluripotent and the trophectodermal cell lineages. These processes are regulated by key transcription factors, likely in cooperation with chromatin modifiers that control histone and DNA methylation. To characterize the spatiotemporal expression of chromatin modifiers in relation to developmental transitions, we performed gene expression profiling of 156 genes in individual oocytes and single blastomeres of developing mouse embryos until the blastocyst stage. More than half of the chromatin modifiers displayed either maternal or zygotic expression. We also detected lineage-specific expression of several modifiers, including Ezh1, Prdm14, Scmh1 and Tet1 underscoring possible roles in cell fate decisions. Members of the SET-domain containing SMYD family showed differential gene expression during preimplantation development. We further observed co-expression of genes with opposing biochemical activities, such as histone methyltransferases and demethylases, suggesting the existence of a dynamic chromatin steady-state during preimplantation development.

Nuclear transfer (NT) technique provides a powerful experimental tool to study the mechanisms of reprogramming processes and to derive NT-embryonic stem (ntES) cells from living or frozen animals. The Piezo-driven direct microinjection NT method has proved to be a valid technique to clone mice and other species. In addition, this method has been broadly used as a versatile tool for many fields of mouse micromanipulation. This chapter describes the "one step method" protocol of nuclear transfer in mouse, which combines injection of a donor cell nucleus and enucleation of MII metaphase in a single manipulation procedure. This protocol describes the isolation and collection of oocytes, treatment of donor cells, visualization of spindle-chromosomal complex, direct injection and enucleation, activation of reconstructed embryos and their in vitro culture and transfer into pseudopregnant mice.

Recently, reprogramming of somatic cells from a differentiated to pluripotent state by overexpression of specific external transcription factors has been accomplished. It has been widely speculated that an undifferentiated state may make donor cells more efficient for nuclear transfer. To test this hypothesis, we derived induced pluripotent stem cells (iPS cells) from several somatic cell lines: mouse embryonic fibroblast (MEF), adult tail tip fibroblast (TTF), and brain neural stem cells (NSCs). Three dimensional (3D)-fluorescent in situ hybridization (FISH) and quantitative-FISH (Q-FISH) were then used to evaluate constitutive (pericentric and telomeric) heterochromatin organization in these iPS cells and in their parental differentiated cells. Here, we show that important nuclear remodeling and telomeres rejuvenation occur in these iPS cells regardless of their parental origin. When we used these cells as donors for nuclear transfer, we produced live-born cloned mice at much higher rates with the iPS-induced cells than with the parental cell lines. Interestingly, we noticed that developmental potential after nuclear transfer could be correlated with telomere length of the donor cells. Altogether, our findings suggest that constitutive heterochromatin organization from differentiated somatic cells can be reprogrammed to the pluripotent state by induction of iPS cells, which in turn support nuclear transfer procedure quite efficiently.

Successful generation of cloned mice using nuclear transfer from induced pluripotent stem cells

Nowadays, a critical question in the epigenetic field is how chromatin is organised within the cell nucleus and how it affects gene expression. In this study we hypothesise that nuclear structure could be involved in the control of gene expression during early rabbit embryonic development. We focused on pericentric/centric heterochromatin, a peculiar region within nuclei known to form higher-order chromatin structures. We therefore performed immunostaining of associated proteins (HP1 and CENP) as well as FISH (fluorescent in situ hybridization) with probes corresponding to these genomic regions. Fertilized embryos were collected from New Zealand white rabbit female, cultured in vitro and fixed at different developmental stages up to 16-cell. Nuclear transfer of rabbit fetal skin fibroblasts was performed by electrofusion and electroactivation (Chesne et al., 2002). Immunostaining experiments were performed as previously described (Martin et al., 2006) and FISH experiments were performed using DNA probes specific to Rsat I and Rsat II sequences (Ekes et al., 2004) according to our published protocol (Maalouf et al., 2010). Three-dimensional images were acquired by confocal laser microscopy (each group included 20 to 40 embryos) and then analysed through automated 3D image processes (according to Ballester et al., 2008; Pichugin et al., 2010). Briefly, after segmentation of the nuclei, immunostaining signals were analysed using an automatic threshold segmentation procedure and FISH spots distributions were analysed with the eroded volume fraction (EVF) method. In in vivo fertilized rabbit embryos, we observed that during the 1- and 2-cell cycles pericentric heterochromatin was dispersed throughout the embryonic nuclei. Large foci of pericentric heterochromatin then progressively appeared at the 4-cell stage and increased in frequency and size by the 8/16-cell stage [∼75% of the embryos, n = 40; i.e. the stage of major transcriptional activation also called MET (maternal to embryonic-transition)]. This suggested that dramatic aggregation of pericentric heterochromatin is associated to the onset of transcription. Interestingly, after nuclear transfer, the donor nuclear organisation was quickly reverted into an early embryonic dispersed form in most 1-cell embryos (80%, n = 35). Subsequently, a somatic-like nuclear reorganization with large heterochromatin foci was re-established in 86% of the 4-cell stage cloned embryos (n = 35). This pattern was similar to the one observed in fertilized embryos at MET; it also appeared one stage earlier (4-cell vs 8-cell). Coincidently, most cloned embryos stopped developing at that stage (85% of cleavage until the 4-cell but only 38% at morula, n = 173). Together, the results suggested that inadequate spatio-temporal reorganization of the donor nucleus could be deleterious during early development in rabbit clones. The present work was supported by INRA « Jeune Equipe » funding, the European CLONET (MRTN-CT-2006-035468) grant and the Ambassade de France en Chine.