BackgroundDue to the critical condition and poor immunity of patients, the intensive care unit (ICU) has always been the main hospital source of multidrug-resistant bacteria. In recent years, with the large-scale use of antibiotics, the detection rate and mortality of carbapenem-resistant Klebsiella pneumoniae (CRKP) have gradually increased. This study explores the molecular characteristics and prevalence of CRKP isolated from the ICU ward of a tertiary hospital in China.MethodsA total of 51 non-duplicated CRKP samples isolated from the ICU were collected from July 2018–July 2020. The enzyme production of the strains was preliminarily screened by carbapenemase phenotypic test, and drug-resistant and virulence genes were detected by PCR. The transferability of plasmid was verified by conjugation test. The minimal inhibitory concentration (MIC) was determined by microbroth dilution method and genetic diversity was detected by multilocus sequence typing and pulsed-field gel electrophoresis.ResultsblaKPC-2 was the only carbapenemase detected. The major virulence genes were uge (100%), mrkD (94.1%), kpn (94.1%), and fim-H (72.5%), while wcag, ironB, alls and magA genes were not detected. One sequence type ST1373 strain, hypervirulent K. pneumoniae (hvKP), was detected. CRKP strains were highly resistant to quinolones, cephalosporins, aminoglycosides, and polymyxin, but susceptive to tigecycline and ceftazidime–avibactam. The success rate of conjugation was 12.2%, indicating the horizontal transfer of blaKPC-2. Homology analysis showed that there was a clonal transmission of ST11 CRKP in the ICU of our hospital.ConclusionThe present study showed the outbreak and dissemination in ICU were caused by ST11 CRKP, which were KPC-2 producers, and simultaneously, also carried some virulence genes. ST11 CRKP persisted in the ward for a long time and spread among different areas. Due to the widespread dispersal of the transferable blaKPC-2 plasmid, the hospital should promptly adopt effective surveillance and strict infection control strategies to prevent the further spread of CRKP. Ceftazidime–avibactam showed high effectiveness against CRKP and could be used for the treatment of ICU infections.

No abstract

Graph neural networks (GNN) have shown great success in learning from graph-structured data. They are widely used in various applications, such as recommendation, fraud detection, and search. In these domains, the graphs are typically large, containing hundreds of millions or billions of nodes. To tackle this challenge, we develop DistDGLv2, a system that extends DistDGL for training GNNs in a mini-batch fashion, using distributed hybrid CPU/GPU training to scale to large graphs. DistDGLv2 places graph data in distributed CPU memory and performs mini-batch computation in GPUs. Dist-DGLv2 distributes the graph and its associated data (initial features) across the machines and uses this distribution to derive a computational decomposition by following an ownercompute rule. DistDGLv2 follows a synchronous training approach and allows ego-networks forming mini-batches to include non-local nodes. To minimize the overheads associated with distributed computations, DistDGLv2 uses a multi-level graph partitioning algorithm with min-edge cut along with multiple balancing constraints. This localizes computation in both machine level and GPU level and statically balance the computations. DistDGLv2 deploys an asynchronous minibatch generation pipeline that makes all computation and data access asynchronous to fully utilize all hardware (CPU, GPU, network, PCIe). The combination allows DistDGLv2 to train high-quality models while achieving high parallel efficiency and memory scalability. We demonstrate DistDGLv2 on various GNN workloads. Our results show that DistDGLv2 achieves 2 − 3× speedup over DistDGL and 18× speedup over Euler. It takes only 5 − 10 seconds to complete an epoch on graphs with 100s millions of nodes on a cluster with 64 GPUs.