Cloning and heterologous expression of pro-2127,a gene encoding cold-active protease from Pseudoalteromonas sp.QI-1

The psychrotropic bacterium, Pseudoalteromonas sp. QI-I, which produces extracellular cold-active protease, was isolated from Antarctic seawater. The genomic DNA of this bacterium was used to construct a plasmid genomic library with the goal of screening cold-active protease genes. Gene pro-2127 with an open reading frame of 2127 bp encoding protease PRO-2127 was cloned and sequenced. Alignment of amino acid sequences suggested that the precursor of PRO-2127 was a member of subfamily S8A, and that it might contain four domains： a signal peptide, an N-terminal prosequence, a catalytic domain and a C-terminal extension. Amino acids Asp185, His244 and Ser425 might form a catalytic triad. PRO-2127 showed some structural features common to psychrophilic enzymes, such as a decrease in Arg residues and the Arg/（Arg＋Lys） ratio. Heterologous expression of pro-2127 in Escherichia coli BL21 （DE3） by pColdlII was also successfully observed in this study.

一种基于随机掩码的低通信量Logistic回归外包训练方案

Logistic回归是一种典型的机器学习模型,因其在疾病诊断、金融预测等许多应用表现优越而受到广泛关注。Logistic回归模型的建立不仅依赖于算法,更依赖于大量有效的训练数据。尽管构建高精度模型并提供预测服务有诸多优点,但用户的敏感信息数据造成隐私问题。因此,该文提出一个新的Logistic回归外包训练方案。在该方案中,用户会预先对私有数据进行处理,并添加随机掩码的数据矩阵上传给聚合器,聚合器将聚合得到的全局训练矩阵上传给云服务器进行训练。该方案在满足数据隐私的安全性需求下具有较高的计算效率和较低的通信开销。

基于BP神经网络模型的饮食方案研究

为推崇健康饮食,满足各类人群的健康饮食需求,文章提出一个基于BP神经网络模型的饮食方案。不同人群对食品的需求千差万别,该文旨在为满足不同人群的多样化需求而为店家提供各式各样的餐饮搭配方案。文章采用BP神经网络模型,对前期问卷所收集的桂林周边饮食需求偏好的数据进行分析,得出了各类人群的乐享健康饮食方案。结果表明,BP神经网络模型在餐饮方面具有较强的适用性。