Jinghan Wang

Acute myeloid leukemia (AML) is a group of hematological malignancies with high heterogeneity. There is an increasing need to improve the risk stratification of AML patients, including those with normal cytogenetics, using molecular biomarkers. Here, we report a metabolomics study that identified a distinct glucose metabolism signature with 400 AML patients and 446 healthy controls. The glucose metabolism signature comprises a panel of 6 serum metabolite markers, which demonstrated prognostic value in cytogenetically normal AML patients. We generated a prognosis risk score (PRS) with 6 metabolite markers for each patient using principal component analysis. A low PRS was able to predict patients with poor survival independently of well-established markers. We further compared the gene expression patterns of AML blast cells between low and high PRS groups, which correlated well to the metabolic pathways involving the 6 metabolite markers, with enhanced glycolysis and tricarboxylic [corrected] acid cycle at gene expression level in low PRS group. In vitro results demonstrated enhanced glycolysis contributed to decreased sensitivity to antileukemic agent arabinofuranosyl cytidine (Ara-C), whereas inhibition of glycolysis suppressed AML cell proliferation and potentiated cytotoxicity of Ara-C. Our study provides strong evidence for the use of serum metabolites and metabolic pathways as novel prognostic markers and potential therapeutic targets for AML.

Binary diffing analysis quantitatively measures the differences between two given binaries and produces fine-grained basic block level matching. It has been widely used to enable different kinds of critical security analysis. However, all existing program analysis and machine learning based techniques suffer from low accuracy, poor scalability, coarse granularity, or require extensive labeled training data to function. In this paper, we propose an unsupervised program-wide code representation learning technique to solve the problem. We rely on both the code semantic information and the program-wide control flow information to generate basic block embeddings. Furthermore, we propose a khop greedy matching algorithm to find the optimal diffing results using the generated block embeddings. We implement a prototype called DEEPBINDIFF and evaluate its effectiveness and efficiency with a large number of binaries. The results show that our tool outperforms the state-of-the-art binary diffing tools by a large margin for both cross-version and cross-optimization-level diffing. A case study for OpenSSL using real-world vulnerabilities further demonstrates the usefulness of our system.

Due to the increasing environmental pollution caused by human activities, environmental remediation has become an important subject for humans and environmental safety. The quest for beneficial pathways to remove organic and inorganic contaminants has been the theme of considerable investigations in the past decade. The easy and quick separation made magnetic solid-phase extraction (MSPE) a popular method for the removal of different pollutants from the environment. Metal–organic frameworks (MOFs) are a class of porous materials best known for their ultrahigh porosity. Moreover, these materials can be easily modified with useful ligands and form various composites with varying characteristics, thus rendering them an ideal candidate as adsorbing agents for MSPE. Herein, research on MSPE, encompassing MOFs as sorbents and Fe3O4 as a magnetic component, is surveyed for environmental applications. Initially, assorted pollutants and their threats to human and environmental safety are introduced with a brief introduction to MOFs and MSPE. Subsequently, the deployment of magnetic MOFs (MMOFs) as sorbents for the removal of various organic and inorganic pollutants from the environment is deliberated, encompassing the outlooks and perspectives of this field.

Summary To forecast the future trend of financial activities through its rules, a convolutional neural network (CNN) is used to forecast stock index. Firstly, a CNN stock index prediction model is constructed, the structural parameter relationship of the CNN model is analyzed, and a CNN model algorithm is implemented. Secondly, the influence of model parameters on prediction results is discussed, and the stock index prediction model based on CNN‐support vector machine (SVM) is established. At last, the empirical analysis is made, and the results show that the two prediction models are feasible and effective. It is concluded that the use of neural networks for financial prediction can deal with the continuous and categorical prediction variables and obtain good prediction results.