Search

Background: Uveal melanoma (UM) is the most frequent ocular neoplasm with a strong metastatic ability. The prognostic value of metastasis-associated genes (MAGs) of UM remains unclear. It is urgent to develop a prognostic score system according to the MAGs of UM.Methods: Unsupervised clustering was used to identify MAGs-based molecular subtypes. Cox methods were utilized to generate a prognostic score system. The prognostic ability of the score system was detected by plotting ROC and survival curves. The immune activity and underlying function were depicted by CIBERSORT GSEA algorithms.Results: Gene cluster analysis determined two MAGs-based subclusters in UM, which were remarkably different in clinical outcomes. A risk score system containing six MAGs (COL11A1, AREG, TIMP3, ADAM12, PRRX1 and GAS1) was set up. We employed ssGSEA to compare immune activity and immunocyte infiltration between the two risk groups. Notch, JAK/STAT and mTOR pathways were greatly enriched in the high-risk group. Furthermore, we observed that knockdown of AREG could inhibit UM proliferation and metastasis by in vitro assays.Conclusion: The MAGs-based subtype and score system in UM can enhance prognosis assessment, and the core system provides valuable reference for clinical decision-making.

Data cleaning is crucial but often laborious in most machine learning (ML) applications. However, task-agnostic data cleaning is sometimes unnecessary if certain inconsistencies in the dirty data will not affect the prediction of ML models to the test points. A test point is certifiably robust for an ML classifier if the prediction remains the same regardless of which (among exponentially many) cleaned dataset it is trained on. In this paper, we study certifiable robustness for the Naive Bayes classifier (NBC) on dirty datasets with missing values. We present (i) a linear time algorithm in the number of entries in the dataset that decides whether a test point is certifiably robust for NBC, (ii) an algorithm that counts for each label, the number of cleaned datasets on which the NBC can be trained to predict that label, and (iii) an efficient optimal algorithm that poisons a clean dataset by inserting the minimum number of missing values such that a test point is not certifiably robust for NBC. We prove that (iv) poisoning a clean dataset such that multiple test points become certifiably non-robust is NP-hard for any dataset with at least three features. Our experiments demonstrate that our algorithms for the decision and data poisoning problems achieve up to $19.5\times$ and $3.06\times$ speed-up over the baseline algorithms across different real-world datasets., Comment: 14 pages, 10 figures

The potential of frequency comb spectroscopy has aroused great interest in generating mid-infrared frequency combs in the integrated photonic setting. However, despite remarkable progress in microresonators and quantum cascade lasers, the availability of suitable mid-IR comb sources remains scarce. Here, we present a new approach for the generation of mid-IR microcombs through cascaded three-wave-mixing. By pumping a CdSiP$_2$ microresonator at 1.55 $μ$m wavelength with a low power continuous wave laser, we generate $χ^{(2)}$ frequency combs at 3.1 $μ$m wavelength, with a span of about 30 nm. We observe ordinary combs states with a line spacing of the free spectral range of the resonator, and combs where the sideband numbers around the pump and half-harmonic alternate, forming staggered patterns of spectral lines. Our scheme for mid-IR microcomb generation is compatible with integrated telecom lasers. Therefore, it has the potential to be used as a simple and fully integrated mid-IR comb source, relying on only one single material.

Most data analytical pipelines often encounter the problem of querying inconsistent data that violate pre-determined integrity constraints. Data cleaning is an extensively studied paradigm that singles out a consistent repair of the inconsistent data. Consistent query answering (CQA) is an alternative approach to data cleaning that asks for all tuples guaranteed to be returned by a given query on all (in most cases, exponentially many) repairs of the inconsistent data. This paper identifies a class of acyclic select-project-join (SPJ) queries for which CQA can be solved via SQL rewriting with a linear time guarantee. Our rewriting method can be viewed as a generalization of Yannakakis's algorithm for acyclic joins to the inconsistent setting. We present LinCQA, a system that can output rewritings in both SQL and non-recursive Datalog rules for every query in this class. We show that LinCQA often outperforms the existing CQA systems on both synthetic and real-world workloads, and in some cases, by orders of magnitude.

Sensitive and accurate detection of DNA methyltransferase (MTase) is conducive to the understanding of the fundamental biological processes related to DNA methylation, clinical disease diagnosis, and drug discovery. Herein, a new fluorescence transducer based on Forster resonance energy transfer (FRET) between the donor upconversion nanoparticles (UCNPs) and the efficient acceptor gold nanorods (AuNRs) for MTase activity analysis and its inhibitor screening is presented. A double-strand DNA linker between UCNPs and AuNRs could be digested by restriction endonuclease HhaI, preventing the FRET process and recovering the upconversion luminescence (UCL) intensity. With the treatment of MTase, the cutting site was disturbed by the methylation of cytosine, blocking the enzyme digestion. The transducer presented here showed an excellent analytical performance toward MTase M.HhaI in the concentration range 0.08~24 U mL−1 with a detection limit of 0.057 U mL−1 calculated according to the UCL intensity changes at 656 nm excited by 980 nm CW laser, which is superior to most of the reported methods. Furthermore, the as-fabricated transducer also demonstrated high testing and screening capability toward enzyme inhibitors’ evaluation. The method takes the advantage of low background fluorescence of UCNPs to improve the accuracy of the measurement, which can be developed as a general strategy for the analysis of various disease-related methyltransferase activity and their corresponding inhibitors, offering a promising strategy for high-performance diagnosis, high-efficient drug exploitation, and treatment effectiveness evaluation.

There is a large body of recent work applying machine learning (ML) techniques to query optimization and query performance prediction in relational database management systems (RDBMSs). However, these works typically ignore the effect of \textit{intra-parallelism} -- a key component used to boost the performance of OLAP queries in practice -- on query performance prediction. In this paper, we take a first step towards filling this gap by studying the problem of \textit{tuning the degree of parallelism (DOP) via ML techniques} in Microsoft SQL Server, a popular commercial RDBMS that allows an individual query to execute using multiple cores. In our study, we cast the problem of DOP tuning as a {\em regression} task, and examine how several popular ML models can help with query performance prediction in a multi-core setting. We explore the design space and perform an extensive experimental study comparing different models against a list of performance metrics, testing how well they generalize in different settings: $(i)$ to queries from the same template, $(ii)$ to queries from a new template, $(iii)$ to instances of different scale, and $(iv)$ to different instances and queries. Our experimental results show that a simple featurization of the input query plan that ignores cost model estimations can accurately predict query performance, capture the speedup trend with respect to the available parallelism, as well as help with automatically choosing an optimal per-query DOP.

Single-cell RNA-seq (scRNA-seq) enables the characterization of transcriptomic profiles at the single-cell resolution with increasingly high throughput. However, it suffers from many sources of technical noises, including insufficient mRNA molecules that lead to excess false zero values, often termed dropouts. Computational approaches have been proposed to recover the biologically meaningful expression by borrowing information from similar cells in the observed dataset. However, these methods suffer oversmoothing and removal of natural cell-to-cell stochasticity in gene expression. Here, we propose the generative adversarial networks for scRNA-seq imputation (scIGANs), which uses generated realistic rather than observed cells to avoid these limitations and the powerless for rare cells. Evaluations based on a variety of simulated and real scRNA-seq datasets demonstrate that scIGANs is effective for dropout imputation and enhancing various downstream analysis. ScIGANs is also scalable and robust to small datasets that have few genes with low expression and/or cell-to-cell variance.

学位の種別: 課程博士, 審査委員会委員 : （主査）東京大学准教授 四本 裕子, 東京大学教授 岡ノ谷 一夫, 東京大学教授 丹野 義彦, 東京大学准教授 本吉 勇, 東京大学准教授 小池 進介

As the increasing concern of driving ranges for electric vehicles, the precision of state of energy (SOE) estimation contributes to the better performance of battery management. In this article, an innovative method based on 5th-order simplex square-radius cubature Kalman filter is developed to achieve the precise and robust estimation for SOE on electric vehicles. Based on second-order equivalent circuit model and particle-swarm-optimization algorithm, the implementation helps the validation of SOE estimation. The max error of SOE estimation under stable condition is less than 3%, and that for dynamic stress test condition is under 4%. Moreover, the robustness is investigated based on diverse deviations on initialization, delivering the future potential applications on embedded system on cloud-controlling.

We develop new models and algorithms for learning the temporal dynamics of the topic polytopes and related geometric objects that arise in topic model based inference. Our model is nonparametric Bayesian and the corresponding inference algorithm is able to discover new topics as the time progresses. By exploiting the connection between the modeling of topic polytope evolution, Beta-Bernoulli process and the Hungarian matching algorithm, our method is shown to be several orders of magnitude faster than existing topic modeling approaches, as demonstrated by experiments working with several million documents in under two dozens of minutes., NeurIPS 2019

Introduction. Although Chinese herbal medicine (CHM) treatment combined with conventional western therapy has been widely used and reported in many clinical trials in China, there is uncertainty about the efficacy of this combination in the treatment of patients after cardiopulmonary resuscitation (CPR). This systematic review aimed to assess whether the risk of mortality has decreased comparing the combination of CHM treatment with conventional western therapy. Methods. To identify relevant studies, the literature search was conducted in Medline, Embase, the Cochrane Library, CBM, CNKI, VIP, and Wanfang database. We included all randomized controlled trials (RCTs) that compared outcomes of patients after CPR taking combination of CHM treatment with those taking just conventional western therapy. Results. This meta-analysis showed that patients randomly assigned to combined CHM treatment group had a statistically significant 23% reduction in mortality compared with those randomly assigned to conventional western therapy group (RR: 0.77; 95% CI: 0.70–0.84). Conclusions. This meta-analysis provides evidence suggesting that a combined CHM therapy is associated with a decreased risk of mortality compared with conventional western therapy in patients after CPR. Further studies are needed to provide more evidence to prove or refute our conclusion and identify reasons for the reduction of mortality.

With the growing number of high-resolution satellite images, the traditional image retrieval method has become a bottleneck in the massive application of high-resolution satellite images because of the low degree of automation. However, there are few studies on the automation of satellite image retrieval. This paper presents an automatic high-resolution satellite image accurate retrieval method based on effective coverage (EC) information, which is used to replace the artificial screening stage in traditional satellite image retrieval tasks. In this method, first, we use a convolutional neural network to extract the EC of each satellite image; then, we use an effective coverage grid set (ECGS) to represent the ECs of all satellite images in the library; finally, the satellite image accurate retrieval algorithm is proposed to complete the process of screening images. The performance evaluation of the method is implemented in three regions: Wuhan, Yanling, and Tangjiashan Lake. The large number of experiments shows that our proposed method can automatically retrieve high-resolution satellite images and significantly improve efficiency.

Based on the relevance of fractal theory, taking Liaoning Province as an example to research the time series of regional economic growth, applied GP algorithm to calculate the correlation dimension and observe whether it has fractal characteristics, correlation dimension is a relatively simple extraction fractal dimension method through experimental data by calculating fractal dimension, has a certain operational, the method lays the foundation for predicting the regional economic growth and judging the economic situation.

International audience; Seven distributors with different configurations are designed and optimized by constructal approach. Their flow distribution performance and energy dissipation are investigated and compared by computational fluid dynamics (CFD) simulation. The reliability of CFD simulation is verified by experiments on the distributor that has all distributing rectangle channels on a plate. The results show that the symmetry of the distributing channels has decisive influence on the performance of flow distribution. Increasing the generations of channel branching will improve the flow distribution uniformity, but on the other hand increase the energy dissipation. Among all the seven constructal distributors, the distributor that has dichotomy configuration, Y-type junctions and straight interconnecting channels, is recommended for its better flow distribution performance and less energy dissipation.