No abstract

Transitional cell carcinoma (TCC) is the most common type of bladder cancer. Here we sequenced the exomes of nine individuals with TCC and screened all the somatically mutated genes in a prevalence set of 88 additional individuals with TCC with different tumor stages and grades. In our study, we discovered a variety of genes previously unknown to be mutated in TCC. Notably, we identified genetic aberrations of the chromatin remodeling genes (UTX, MLL-MLL3, CREBBP-EP300, NCOR1, ARID1A and CHD6) in 59% of our 97 subjects with TCC. Of these genes, we showed UTX to be altered substantially more frequently in tumors of low stages and grades, highlighting its potential role in the classification and diagnosis of bladder cancer. Our results provide an overview of the genetic basis of TCC and suggest that aberration of chromatin regulation might be a hallmark of bladder cancer.

Pyroptosis is critical for macrophages against pathogen infection, but its role and mechanism in cancer cells remain unclear. PD-L1 has been detected in the nucleus with unknown function. Here, we show that PD-L1 switches TNFα-induced apoptosis to pyroptosis in cancer cells, resulting in tumor necrosis. Under hypoxia, p-Stat3 physically interacts with PD-L1 and facilitates its nuclear translocation, enhancing gasdermin C (GSDMC) gene transcription. GSDMC is specifically cleaved by caspase-8 with TNFα treatment, generating a GSDMC N-terminal domain that forms pores on cell membrane and induces pyroptosis. Nuclear PD-L1, caspase-8, and GSDMC are required for macrophage-derived TNFα-induced tumor necrosis in vivo . Moreover, high expression of GSDMC correlates with poor survival. Antibiotic chemotherapy drugs induce pyroptosis in breast cancer. These findings identify a non-immune checkpoint function of PD-L1 and provide an unexpected concept that GSDMC/Caspas-8 mediates non-canonical pyroptosis pathway in cancer cells, causing tumor necrosis.

Bladder cancer is one of the most common cancers worldwide, with transitional cell carcinoma (TCC) being the predominant form. Here we report a genomic analysis of TCC by both whole-genome and whole-exome sequencing of 99 individuals with TCC. Beyond confirming recurrent mutations in genes previously identified as being mutated in TCC, we identified additional altered genes and pathways that were implicated in TCC. Notably, we discovered frequent alterations in STAG2 and ESPL1, two genes involved in the sister chromatid cohesion and segregation (SCCS) process. Furthermore, we also detected a recurrent fusion involving FGFR3 and TACC3, another component of SCCS, by transcriptome sequencing of 42 DNA-sequenced tumors. Overall, 32 of the 99 tumors (32%) harbored genetic alterations in the SCCS process. Our analysis provides evidence that genetic alterations affecting the SCCS process may be involved in bladder tumorigenesis and identifies a new therapeutic possibility for bladder cancer.

IMPORTANCE Laparoscopic distal gastrectomy is accepted as a more effective approach to conventional open distal gastrectomy for early-stage gastric cancer. However, efficacy for locally advanced gastric cancer remains uncertain. OBJECTIVE To compare 3-year disease-free survival for patients with locally advanced gastric cancer after laparoscopic distal gastrectomy or open distal gastrectomy. DESIGN, SETTING, AND PATIENTS The study was a noninferiority, open-label, randomized clinical trial at 14 centers in China. A total of 1056 eligible patients with clinical stage T2, T3, or T4a gastric cancer without bulky nodes or distant metastases were enrolled from September 2012 to December 2014. Final follow-up was on December 31, 2017. INTERVENTIONS Participants were randomized in a 1:1 ratio after stratification by site, age, cancer stage, and histology to undergo either laparoscopic distal gastrectomy (n = 528) or open distal gastrectomy (n = 528) with D2 lymphadenectomy. MAIN OUTCOMES AND MEASURESThe primary end point was 3-year disease-free survival with a noninferiority margin of −10% to compare laparoscopic distal gastrectomy with open distal gastrectomy. Secondary end points of 3-year overall survival and recurrence patterns were tested for superiority. RESULTS Among 1056 patients, 1039 (98.4%; mean age, 56.2 years; 313 [30.1%] women) had surgery (laparoscopic distal gastrectomy [n=519] vs open distal gastrectomy [n=520]), and 999 (94.6%) completed the study. Three-year disease-free survival rate was 76.5% in the laparoscopic distal gastrectomy group and 77.8% in the open distal gastrectomy group, absolute difference of −1.3% and a 1-sided 97.5% CI of −6.5% to ϱ, not crossing the prespecified noninferiority margin. Three-year overall survival rate (laparoscopic distal gastrectomy vs open distal gastrectomy: 83.1% vs 85.2%; adjusted hazard ratio, 1.19; 95% CI, 0.87 to 1.64; P = .28) and cumulative incidence of recurrence over the 3-year period (laparoscopic distal gastrectomy vs open distal gastrectomy: 18.8% vs 16.5%; subhazard ratio, 1.15; 95% CI, 0.86 to 1.54; P = .35) did not significantly differ between laparoscopic distal gastrectomy and open distal gastrectomy groups.CONCLUSIONS AND RELEVANCE Among patients with a preoperative clinical stage indicating locally advanced gastric cancer, laparoscopic distal gastrectomy, compared with open distal gastrectomy, did not result in inferior disease-free survival at 3 years.

Training deep models for lane detection is challenging due to the very subtle and sparse supervisory signals inherent in lane annotations. Without learning from much richer context, these models often fail in challenging scenarios, e.g., severe occlusion, ambiguous lanes, and poor lighting conditions. In this paper, we present a novel knowledge distillation approach, i.e., Self Attention Distillation (SAD), which allows a model to learn from itself and gains substantial improvement without any additional supervision or labels. Specifically, we observe that attention maps extracted from a model trained to a reasonable level would encode rich contextual information. The valuable contextual information can be used as a form of 'free' supervision for further representation learning through performing topdown and layer-wise attention distillation within the network itself. SAD can be easily incorporated in any feedforward convolutional neural networks (CNN) and does not increase the inference time. We validate SAD on three popular lane detection benchmarks (TuSimple, CULane and BDD100K) using lightweight models such as ENet, ResNet-18 and ResNet-34. The lightest model, ENet-SAD, performs comparatively or even surpasses existing algorithms. Notably, ENet-SAD has 20 × fewer parameters and runs 10 × faster compared to the state-of-the-art SCNN [16], while still achieving compelling performance in all benchmarks. Our code is available at https://github. com/cardwing/Codes-for-Lane-Detection.

We propose Stochastic Neural Architecture Search (SNAS), an economical endto-end solution to Neural Architecture Search (NAS) that trains neural operation parameters and architecture distribution parameters in same round of backpropagation, while maintaining the completeness and differentiability of the NAS pipeline. In this work, NAS is reformulated as an optimization problem on parameters of a joint distribution for the search space in a cell. To leverage the gradient information in generic differentiable loss for architecture search, a novel search gradient is proposed. We prove that this search gradient optimizes the same objective as reinforcement-learning-based NAS, but assigns credits to structural decisions more efficiently. This credit assignment is further augmented with locally decomposable reward to enforce a resource-efficient constraint. In experiments on CIFAR-10, SNAS takes fewer epochs to find a cell architecture with state-of-theart accuracy than non-differentiable evolution-based and reinforcement-learningbased NAS, which is also transferable to ImageNet. It is also shown that child networks of SNAS can maintain the validation accuracy in searching, with which attention-based NAS requires parameter retraining to compete, exhibiting potentials to stride towards efficient NAS on big datasets. INTRODUCTIONThe trend to seek for state-of-the-art neural network architecture automatically has been growing since Zoph & Le (2016), given the enormous effort needed in scientific research. Normally, a Neural Architecture Search (NAS) pipeline comprises architecture sampling, parameter learning, architecture validation, credit assignment and search direction update.There are basically three existing frameworks for neural architecture search. Evolution-based NAS like NEAT (Stanley & Miikkulainen, 2002) employs evolution algorithm to simultaneously optimize topology alongside with parameters. However, it takes enormous computational power and could not leverage the efficient gradient back-propagation in deep learning. To achieve the state-of-the-art performance as human-designed architectures, Real et al. ( 2018) takes 3150 GPU days for the whole evolution. Reinforcement-learning-based NAS is end-to-end for gradient back-propagation, among which the most efficient one, ENAS (Pham et al., 2018) learns optimal parameters and architectures together just like NEAT. However, as NAS is modeled as a Markov Decision Process, credits are assigned to structural decisions with temporal-difference (TD) learning (Sutton et al., 1998), whose efficiency and interpretability suffer from delayed rewards (Arjona-Medina et al., 2018). To get rid of the architecture sampling process, DARTS (Liu et al., 2019) proposes deterministic attention on operations to analytically calculate expectation at each layer. After the convergence of the parent network, it removes operations with relatively weak attention. Due to the pervasive non-linearity in neural operations, it introduces untractable bias to the loss function. This bias causes ...

Abstract. State-of-the-art person re-identification methods seek robust person matching through combining various feature types. Often, these features are implicitly assigned with a single vector of global weights, which are assumed to be universally good for all individuals, independent to their different appearances. In this study, we show that certain features play more important role than others under different circumstances. Consequently, we propose a novel unsupervised approach for learning a bottom-up feature importance, so features extracted from different individuals are weighted adaptively driven by their unique and inherent appearance attributes. Extensive experiments on two public datasets demonstrate that attribute-sensitive feature importance facilitates more accurate person matching when it is fused together with global weights obtained using existing methods.