The central building block of convolutional neural networks (CNNs) is the convolution operator, which enables networks to construct informative features by fusing both spatial and channel-wise information within local receptive fields at each layer. A broad range of prior research has investigated the spatial component of this relationship, seeking to strengthen the representational power of a CNN by enhancing the quality of spatial encodings throughout its feature hierarchy. In this work, we focus instead on the channel relationship and propose a novel architectural unit, which we term the "Squeeze-and-Excitation" (SE) block, that adaptively recalibrates channel-wise feature responses by explicitly modelling interdependencies between channels. We show that these blocks can be stacked together to form SENet architectures that generalise extremely effectively across different datasets. We further demonstrate that SE blocks bring significant improvements in performance for existing state-of-the-art CNNs at slight additional computational cost. Squeeze-and-Excitation Networks formed the foundation of our ILSVRC 2017 classification submission which won first place and reduced the top-5 error to 2.251%, surpassing the winning entry of 2016 by a relative improvement of ∼25%. Models and code are available at https://github.com/hujie-frank/SENet.

5-methylcytosine (5mC) in DNA plays an important role in gene expression, genomic imprinting, and suppression of transposable elements. 5mC can be converted to 5-hydroxymethylcytosine (5hmC) by the Tet proteins. Here we show that, in addition to 5hmC, the Tet proteins can generate 5-formylcytosine (5fC) and 5-carboxylcytosine (5caC) from 5mC in an enzymatic activity-dependent manner. Furthermore, we reveal the presence of 5fC and 5caC in genomic DNA of mouse ES cells and mouse organs. The genomic content of 5hmC, 5fC, and 5caC can be increased or reduced through overexpression or depletion of Tet proteins. Thus, we identify two new cytosine derivatives in genomic DNA as the products of Tet proteins. Our study raises the possibility that DNA demethylation may occur through Tet-catalyzed oxidation followed by decarboxylation.

In this paper, we introduce a new large-scale face dataset named VGGFace2. The dataset contains 3.31 million images of 9131 subjects, with an average of 362.6 images for each subject. Images are downloaded from Google Image Search and have large variations in pose, age, illumination, ethnicity and profession (e.g. actors, athletes, politicians).The dataset was collected with three goals in mind: (i) to have both a large number of identities and also a large number of images for each identity; (ii) to cover a large range of pose, age and ethnicity; and (iii) to minimise the label noise. We describe how the dataset was collected, in particular the automated and manual filtering stages to ensure a high accuracy for the images of each identity.To assess face recognition performance using the new dataset, we train ResNet-50 (with and without Squeeze-and-Excitation blocks) Convolutional Neural Networks on VG-GFace2, on MS-Celeb-1M, and on their union, and show that training on VGGFace2 leads to improved recognition performance over pose and age. Finally, using the models trained on these datasets, we demonstrate state-of-the-art performance on the face recognition of IJB datasets, exceeding the previous state-of-the-art by a large margin. The dataset and models are publicly available 1 .

Acoustic scene classification (ASC) is one of the most popular problems in the field of machine listening. The objective of this problem is to classify an audio clip into one of the predefined scenes using only the audio data. This problem has considerably progressed over the years in the different editions of DCASE. It usually has several subtasks that allow to tackle this problem with different approaches. The subtask presented in this report corresponds to a ASC problem that is constrained by the complexity of the model as well as having audio recorded from different devices, known as mismatch devices (real and simulated). The work presented in this report follows the research line carried out by the team in previous years. Specifically, a system based on two steps is proposed: a two-dimensional representation of the audio using the Gamamtone filter bank and a convolutional neural network using squeeze-excitation techniques. The presented system outperforms the baseline by about 17 percentage points.

Purpose EMT has been associated with metastatic spread and EGFR inhibitor resistance. We developed and validated a robust 76-gene EMT signature using gene expression profiles from four platforms using NSCLC cell lines and patients treated in the BATTLE study. Methods We conducted an integrated gene expression, proteomic, and drug response analysis using cell lines and tumors from NSCLC patients. A 76-gene EMT signature was developed and validated using gene expression profiles from four microarray platforms of NSCLC cell lines and patients treated in the BATTLE (Biomarker-integrated Approaches of Targeted Therapy for Lung Cancer Elimination) study, and potential therapeutic targets associated with EMT were identified. Results Compared with epithelial cells, mesenchymal cells demonstrated significantly greater resistance to EGFR and PI3K/Akt pathway inhibitors, independent of EGFR mutation status, but more sensitivity to certain chemotherapies. Mesenchymal cells also expressed increased levels of the receptor tyrosine kinase Axl and showed a trend towards greater sensitivity to the Axl inhibitor SGI-7079, while the combination of SGI-7079 with erlotinib reversed erlotinib resistance in mesenchymal lines expressing Axl and in a xenograft model of mesenchymal NSCLC. In NSCLC patients, the EMT signature predicted 8-week disease control in patients receiving erlotinib, but not other therapies. Conclusion We have developed a robust EMT signature that predicts resistance to EGFR and PI3K/Akt inhibitors, highlights different patterns of drug responsiveness for epithelial and mesenchymal cells, and identifies Axl as a potential therapeutic target for overcoming EGFR inhibitor resistance associated with the mesenchymal phenotype

The highly complex structure of the human brain is strongly shaped by genetic influences1. Subcortical brain regions form circuits with cortical areas to coordinate movement2, learning, memory3 and motivation4, and altered circuits can lead to abnormal behaviour and disease2. To investigate how common genetic variants affect the structure of these brain regions, here we conduct genome-wide association studies of the volumes of seven subcortical regions and the intracranial volume derived from magnetic resonance images of 30,717 individuals from 50 cohorts. We identify five novel genetic variants influencing the volumes of the putamen and caudate nucleus. We also find stronger evidence for three loci with previously established influences on hippocampal volume5 and intracranial volume6. These variants show specific volumetric effects on brain structures rather than global effects across structures. The strongest effects were found for the putamen, where a novel intergenic locus with replicable influence on volume (rs945270; P = 1.08 × 10−33; 0.52% variance explained) showed evidence of altering the expression of the KTN1 gene in both brain and blood tissue. Variants influencing putamen volume clustered near developmental genes that regulate apoptosis, axon guidance and vesicle transport. Identification of these genetic variants provides insight into the causes of variability inhuman brain development, and may help to determine mechanisms of neuropsychiatric dysfunction.

The outbreak of the novel coronavirus in China (SARS-CoV-2) that began in December 2019 presents a significant and urgent threat to global health. This study was conducted to provide the international community with a deeper understanding of this new infectious disease. Epidemiological, clinical features, laboratory findings, radiological characteristics, treatment, and clinical outcomes of 135 patients in northeast Chongqing were collected and analyzed in this study. A total of 135 hospitalized patients with COVID-19 were enrolled. The median age was 47 years (interquartile range, 36-55), and there was no significant gender difference (53.3% men). The majority of patients had contact with people from the Wuhan area. Fortythree (31.9%) patients had underlying disease, primarily hypertension (13 [9.6%]), diabetes (12 [8.9%]), cardiovascular disease (7 [5.2%]), and malignancy (4 [3.0%]). Common symptoms included fever (120 [88.9%]), cough (102 [76.5%]), and fatigue (44 [32.5%]). Chest computed tomography scans showed bilateral patchy shadows or ground glass opacity in the lungs of all the patients. All patients received antiviral therapy (135 [100%]) (Kaletra and interferon were both used), antibacterial therapy (59 [43.7%]), and corticosteroids (36 [26.7%]). In addition, many patients received traditional Chinese medicine (TCM) (124 [91.8%]). It is suggested that patients should receive Kaletra early and should be treated by a combination of Western and Chinese medicines. Compared to the mild cases, the severe ones had lower Suxin Wan, Yi Xiang, and Wei Fang are the co-first authors.

SUMMARY Human embryonic stem cells (hESCs) share an identical genome with lineage-committed cells, yet possess the remarkable properties of self-renewal and pluripotency. The diverse cellular properties in different cells have been attributed to their distinct epigenomes, but how much epigenomes differ remains unclear. Here, we report that epigenomic landscapes in hESC and lineage committed cells are drastically different. By comparing the chromatin modification profiles and DNA methylomes in hESCs and primary fibroblasts, we find that nearly one-third of the genome differs in chromatin structure. Most changes arise from dramatic redistributions of repressive H3K9me3 and H3K27me3 marks, which form blocks that significantly expand in fibroblasts. A large number of potential regulatory sequences also exhibit a high degree of dynamics in chromatin modifications and DNA methylation. Additionally, we observe novel, context-dependent relationships between DNA methylation and chromatin modifications. Our results provide new insights into epigenetic mechanisms underlying properties of pluripotency and cell-fate commitment.