蛋白质二级结构预测方法研究

为提高蛋白质二级结构预测精度,提出一种新的网络模型和编码方法。首先利用基因表达式编程（GEP）的全局搜索能力同时进化设计神经网络的结构和连接权;其次,对神经网络输入层编码进行了改进,添加了氨基酸残基所处的疏水环境。用PDBSelect25中的36条蛋白质共6 122个残基进行测试,结果表明提出的网络模型和编码方法能有效提高蛋白质二级结构预测的精度。     

基于卷积长短时记忆神经网络的蛋白质二级结构预测

鉴于不同类型氨基酸的相互作用对蛋白质结构预测的影响不同,文中融合卷积神经网络和长短时记忆神经网络模型,提出卷积长短时记忆神经网络,并应用到蛋白质8类二级结构的预测中.首先基于氨基酸序列的类别信息和氨基酸结构的进化信息表示蛋白质序列,并采用卷积提取氨基酸残基之间的局部相关特征,然后利用双向长短时记忆神经网络提取蛋白质序列内部残基之间的远程相互作用,最后将提取的蛋白质的局部相关特征和远程相互作用用于蛋白质8类二级结构的预测.实验表明,相比基准方法,文中模型提高8类二级结构预测的精度,并具有良好的可扩展性.     

基于GEP-BP网络集成的蛋白质二级结构预测方法研究*

为提高蛋白质二级结构预测的精度,提出了一种基于GEP-BP网络集成的两层结构预测模型。首先利用基因表达式编程（GEP）的全局搜索能力同时进化设计BP网络的结构和连接权,并将进化最后一代的个体用BP算法进一步训练学习,然后采用组合方法将部分个体集成构成模型的第一层;根据神经网络输出之间具有相关性,用第二层网络对第一层的预测结果进行精炼。用PDBSelect25中的36条蛋白质共6 122个残基进行测试,结果表明提出的模型能有效预测蛋白质二级结构,将预测精度提高到73.02%。     

基于级联神经网络的蛋白质二级结构预测

为提高蛋白质二级结构预测的精度,提出一种由两层网络构成的级联神经网络模型。第1层网络采用具有差异度的5个子网构成的网络模型,对第2层网络的输入编码进行改进。对PDBSelect25中的36条蛋白质共6 122个残基进行测试,结果表明,该模型能有效预测蛋白质二级结构,其预测精度分别比SNN, DSC, PREDSATOR方法提高5.31%, 1.21%和0.92%,平均预测精度提高到69.61%。     

基于关联规则与遗传算法的蛋白质二级结构预测

文章通过建立蛋白质二级结构预测的数学模型,运用挖掘与遗传算法相结合的关联规则技术对蛋白质二级结构进行预测,设计并实现了该原型系统。实验表明,该文所采用的基于蛋白质氨基酸疏水性周期规律的预测模型方法较其它相关的二级结构预测方法有较好的准确性、有效性与可行性。     

基于径向基神经网络的蛋白质二级结构预测

为提高蛋白质二级结构预测的精确度,提出并构建精确的径向基神经网络、广义回归神经网络,并基于5位编码和Profile编码,采用不同大小的滑动窗口,利用交叉检证法构建多个径向基网络预测器,分别对蛋白质二级结构进行预测,得到了较好的实验结果,其中aveQ3提高到70.96%。结果表明,径向基神经网络模型能有效提高预测精确度,也证明了实验方法的有效性和可行性。     

合成金字塔预测模型中内含的改进型CBA预测方法*

蛋白质二级结构预测问题,是生物信息学领域中最为重要的任务之一,历经三十多年的研究,已取得了一些进展,尤其是近来集成预测模型与混合预测模型的引入,为预测精度带来了一定程度的提高,然而其离从二级结构推导三级结构的目标,仍然存在很大差距。为了有效提高蛋白质二级结构预测精度,以KDTICM理论的扩展性研究与KDD*模型为基础, 使用基于KDD*模型的关联分析蛋白质二级结构预测方法KAAPRO,提出一种基于支持度与可信度的复杂距离度量的CBA(classification based on association)     

结合蛋白质二级结构信息预测蛋白质空间结构中的二硫键*

在蛋白质空间结构预测中,二硫键的确定可以大大减少蛋白质构象的搜索空间。为提高二硫键预测的准确率,对形成二硫键的半胱氨酸及其周围的氨基酸残基在蛋白质二级结构形成上的偏性进行了分析,并提出将蛋白质二级结构信息加入到BP神经网络预测模型的输入编码信息中。研究对象为从SWISS-PROT数据库中选取的252条蛋白质序列,随机均分4组,对预测准确率进行4-交叉验证。各项准确率均比未加入蛋白质二级结构信息前,有明显提高。结果表明,结合蛋白质二级结构信息的编码方式是可行且有效的。     

基于距离矩阵灰度图的蛋白质二级结构类型预测*

蛋白质二级结构预测在蛋白质结构预测中具有很重要的作用。基于伪氨基酸成分表示蛋白质的方法,能提高蛋白质结构和功能预测的成功率,利用蛋白质距离矩阵灰度图,基于几何矩提出了一种伪氨基酸构造方法,结合氨基酸的成分对蛋白质二级结构类型进行预测,通过国际公认的Jackknife检验方法显示预测成功率达到95.10%,比其他方法高出许多,说明此方法具有有效的分类效果。     

元胞自动机图的蛋白质二级结构类型预测

蛋白质结构预测是后基因组时代的一项重要任务,蛋白质二级结构预测是蛋白质结构预测的关键步骤。利用氨基酸数字编码模型生成蛋白质序列的元胞自动机图（Cellular Automata Image,CAI）,提出了一种基于灰度共生矩阵（Gray Level Co-occurrence Matrix,GLCM）提取纹理图像特征的方法。用扩大的协方差算法进行预测,仿真结果显示有较好的分类效果,Jackknife检验的预测成功率达到94.61％。     

Gaussian conditional random fields extended for directed graphs

For many real-world applications, structured regression is commonly used for predicting output variables that have some internal structure. Gaussian conditional random fields (GCRF) are a widely used type of structured regression model that incorporates the outputs of unstructured predictors and the correlation between objects in order to achieve higher accuracy. However, applications of this model are limited to objects that are symmetrically correlated, while interaction between objects is asymmetric in many cases. In this work we propose a new model, called Directed Gaussian conditional random fields (DirGCRF), which extends GCRF to allow modeling asymmetric relationships (e.g. friendship, influence, love, solidarity, etc.). The DirGCRF models the response variable as a function of both the outputs of unstructured predictors and the asymmetric structure. The effectiveness of the proposed model is characterized on six types of synthetic datasets and four real-world applications where DirGCRF was consistently more accurate than the standard GCRF model and baseline unstructured models.     

搜索引擎日志中“N+V”和“N1+N2+V”型短语自动识别

正确识别搜索引擎日志中的短语,对搜索引擎用短语词典构建和提高搜索引擎性能具有重要的作用。该文提出一种应用条件随机场实现对搜狗日志语料中“N+V”和“N₁+N₂+V”型短语自动识别的方法。模型的特征集包含词、词性和词语长度。由人工设计候选特征集,从中选择有效的特征构成特征模板,训练生成用于短语自动识别的条件随机场模型。封闭测试和开放测试的实验结果表明,模型能够实现对这两种短语的有效识别。     

Stochastic inverse method to identify parameter random fields in a structure

The parameters in a structure such as geometric and material properties are generally uncertain due to manufacturing tolerance, wear, fatigue and material irregularity. Such parameters are random fields because the uncertain properties vary along the spatial domain of a structure. Since the parameter uncertainties in a structure result in the uncertainty of the structural dynamic behavior, they need to be identified accurately for structural analysis or design. In order to identify the random fields of geometric parameters, the parameters can be measured directly using a 3-dimensional coordinate measuring machine. However, it is often very expensive to measure them directly. It is even impossible to directly measure some parameters such as density and Young’s modulus. For that case, the parameter random fields should be identified from measurable response data samples. In this paper, a stochastic inverse method to identify parameter random fields in a structure using modal data is proposed. The proposed method consists of the following three steps: (i) obtaining realizations of the parameter random field from modal data samples by solving an optimization problem, (ii) obtaining the deterministic terms in the Karhunen-Loève expansion by solving an eigenvalue problem and (iii) estimating the distributions of random variables in the Karhunen-Loève expansion using a maximum likelihood estimation method with kernel density.     

基于增强约束条件随机场的Web对象信息抽取

线性链条件随机场模型难以处理Web对象与各个标注属性之间的特征关系,为解决此问题,提出一种增强约束条件随机场模型。通过将约束条件引入推理过程,改进线性链条件随机场模型的Viterbi算法;运用最大间隔理论的思想训练条件随机场模型,提高模型标注的正确率;将该模型与条件随机场模型及层次条件随机场模型进行对比。实验结果表明该模型能在提高标注正确率的基础上有效地解决Web对象信息抽取问题。     

Generation of two-dimensionally variable saturated hydraulic conductivity fields: Model theory,verification and computer program

The theory for the generation method of spatially variable K_s fields, using a set of scaling factors of log-normally distributed random field with two-dimensional correlation structure, is compiled from different sources. It is developed further by derivation of computational formulas of model parameters for a general case. The user of the method needs to define the mean, coefficient of variation and correlation length of the K_s fields. Both isotropic and anisotropic fields can be generated. The method and the computer program, which is provided for readers, are verified by analyzing example generations carried out by set of different parameter values. Since the materializations of the set values in generated fields were found satisfactorily accurate and their distributions cohered the model structure, the method and the computer program can be considered a useful and reliable tool for simulation of spatially variable hydraulic conductivity fields. However, the way of the method treating the covariance structure of the random vector limits the field size or the number of nodes where the K_s values are to be generated.     

Displaying Random Fields

This paper presents an algorithm for random fields generation. The main idea of the paper is an improvement of the recursive technique presented by A. Fournier, D. Fussel and L. Carpenter in [4]. In order to ensure the continuity constraints on the boundaries of the cells generated at different stages of the algorithm, we show that it is possible to store the boundary values using a dynamic data structure. Random fields have been successfully used to model height fields or to display synthetic textures. In the second part of this paper we investigate the perspective mapping and clipping stages of random fields generation and display. At the end of the paper, some synthetic skies displayed using this technique are presented.     

基于层叠条件随机场的CFN自动标注研究

提出了一种基于层叠条件随机场的CFN自动标注方法。该方法在低层条件随机场模型中解决了框架元素的识别,将识别结果传递到上层短语类型识别的条件随机场模型,再将识别结果传递到上层句法功能识别的条件随机场模型,其低层模型为上层模型提供决策支持。实验选用CFN中"陈述"框架下的句子库,实现了基于层叠条件随机场CFN自动标注的原型系统。     

基于约束条件随机场的Web对象属性标注

条件随机场模型是目前处理We b对象属性标注问题的最佳统计模型。为解决条件随机场模型不能充分利用We b对象和属性标签之间的特征关系这一问题,提出了一种增强约束条件随机场模型。借鉴最大间隔的思想,在原有条件随机场模型中增加约束条件和增强因子以提高模型标注正确率。使用最大似然参数估计方法估计模型特征函数的权重参数,并用Viterbi算法进行预测。在数据集中引入验证集的概念,以获得最优增强因子。实验结果表明,该模型有效地提高了We b对象属性标注正确率。     

Hidden Markov random field model selection criteria based on mean field-like approximations

Hidden Markov random fields appear naturally in problems such as image segmentation, where an unknown class assignment has to be estimated from the observations at each pixel. Choosing the probabilistic model that best accounts for the observations is an important first step for the quality of the subsequent estimation and analysis. A commonly used selection criterion is the Bayesian Information Criterion (BIC) of Schwarz (1978), but for hidden Markov random fields, its exact computation is not tractable due to the dependence structure induced by the Markov model. We propose approximations of BIC based on the mean field principle of statistical physics. The mean field theory provides approximations of Markov random fields by systems of independent variables leading to tractable computations. Using this principle, we first derive a class of criteria by approximating the Markov distribution in the usual BIC expression as a penalized likelihood. We then rewrite BIC in terms of normalizing constants, also called partition functions, instead of Markov distributions. It enables us to use finer mean field approximations and to derive other criteria using optimal lower bounds for the normalizing constants. To illustrate the performance of our partition function-based approximation of BIC as a model selection criterion, we focus on the preliminary issue of choosing the number of classes before the segmentation task. Experiments on simulated and real data point out our criterion as promising: It takes spatial information into account through the Markov model and improves the results obtained with BIC for independent mixture models.     

基于条件随机场的蒙古语词切分研究

词干和构形附加成分是蒙古语词的组成成分,在构形附加成分中包含着数、格、体、时等大量语法信息。利用这些语法信息有助于使用计算机对蒙古语进行有效处理。蒙古语词在结构上表现为一个整体,为了利用其中的语法信息需要识别出词干和各构形附加成分。通过分析蒙古语词的构形特点,提出一种有效的蒙古语词标注方法,并基于条件随机场模型构建了一个实用的蒙古语词切分系统。实验表明该系统的词切分准确率比现有蒙古语词切分系统的准确率有较大提高,达到了0.992。