基于混合选择的GEP算法及其在喹诺酮羧酸类HIV-1整合酶抑制剂活性的QSAR研究中的应用

提出1种混合使用模拟退火和竞赛规则选择算子的改进GEP算法-GEPMS算法。以E-Dragon软件计算、经RM算法筛选得到的7个RDF描述符作为自变量,以抗HIV-1活性IC_(50)值作为因变量,基于GEPMS算法建立关于48种喹诺酮羧酸类化合物的HIV-1整合酶抑制剂活性的QSAR模型。与GEP、GEPSA和v-SVM算法建立的QSAR模型进行比较,本文模型、GEP、GEPSA和v-SVM模型对训练集的计算结果,决定系数R~2分别为0.9667、0.9624、0.9348和0.9711,对验证集的预测结果R~2则分别为0.9565、0.8974、0.9124和0.7656,表明本文的GEPMS模型具有最佳的泛化能力,算法的改进效果明显。     

新非核苷喹诺酮类HIV-1逆转录酶抑制剂的CoMFA研究

目的：应用比较分子力场法(COMFA)研究一系列喹诺酮类对HIV-1逆转录酶抑制活性的三维定量构效关系,为进一步抗HIV药物设计提供理论依据。方法和结果：在研究的29个化合物中,用比较分子力场法得到一个CoMFA模型,交叉验证系数为q~2=0.556,具有较高的预测能力及合理性,非交叉验证模型相关系数分别为r~2=0.998,标准偏差SE=0.044,F= 401.038;结论：此模型对设计和预测高活性的喹诺酮类HIV-1逆转录酶抑制活性的化合物有一定可靠性。     

R基团搜索技术用于硫代氨基甲酸酯类非核苷类逆转录酶抑制剂的分子设计

硫代氨基甲酸酯（TCS）被确认为一种新的非核苷类HIV-1逆转录酶抑制剂。本文采用基于R基团搜索技术的Topomer Co MFA方法。实验一对111个硫代氨基甲酸酯衍生物分子进行了三维定量构效关系分析,得到3D-QSAR模型的q^2为0.616,r^2为0.751,实验二对前60个分子进行同样的分析,得到q^2为0.777,r^2为0.913。两次实验结果表明所建立的模型在统计上都有较好的稳定性和预测能力,但相比之下,实验二更有利于抗艾滋病药物的设计。     

喹诺酮酸类HIV-1整合酶链转移抑制剂的2D-QSAR研究

HIV-1整合酶(integrase,IN)是病毒复制过程的关键酶,已被证实是开发抗HIV-1药物的一个理想靶标。针对48个喹诺酮酸类整合酶链转移抑制剂(integrasestrandtransfer inhibitors, INSTIs),利用遗传函数逼近法(genetic function approximation, GFA)构建10个抑制活性与优选的分子结构描述符之间的二维定量构效关系(2D-QSAR)模型,从中优选出最佳的模型并对其进行验证,据此探究影响抑制剂生物活性的主要分子微观结构因素,希冀为其进一步结构优化提供理论指导。所建立的最优2D-QSAR模型的非交叉验证相关系数R~2为0.8903,交叉验证相关系数Q~2为0.8213,表明该模型具有较高的预测能力和明显的统计学意义。该研究表明,喹诺酮酸类INSTIs的生物活性主要受Jurs_RPCG、Shadow_nu、BIC、ALogP、Dipole_X以及Dipole_Y描述符的影响,为其进一步结构修饰,开发高效抗HIV-1药物奠定了理论基础。     

苯酚及其衍生物对水生梨形四膜虫急性毒性的QSTR研究

选取了258个苯酚及其衍生物对水生梨形四膜虫的毒性数据,选择7个分子描述符作为建模的结构参数,开展了以QSTR方法建立苯酚及其衍生物毒性模型的研究。首先,运用稳健诊断方法(Robust Diagnostic Method)剔除奇异样本,然后采用球型排除算法(Sphere-exclusion Algorithms)合理划分样本,继而分别采用多元线性回归(Multiple Linear Regression,MLR)、偏最小二乘(Partial Least Squares,PLS)、BP(Error Back-Propagation,BP)神经网络3种方法进行定量构效关系研究,并对外部验证集采用共识建模方法(Consensus Modeling Method),从而提高了模型的预测能力。研究结果表明,所建模型均具有较好的预测能力和稳定性,且与MLR、PLS模型相比,BP神经网络模型性能略胜一筹,即非线性模型比线性模型性能优越。但是BP神经网络建立的模型不能直接给出直观的数学模型和公式,而MLR、PLS模型更为简单明了。     

分子表面随机采样分析用于HIV整合酶抑制剂的QSAR研究

采用分子表面随机采样分析对26个香豆素类衍生物抗艾滋病药物进行定量构效关系研究。运用多元线性回归(multiple linear regression,MLR)建模,同时采用内部及外部双重验证的办法对所得模型稳定性能进行深入分析和检验。MLR建模的复相关系数(_(cum)~2)、留一法(1eave-one-out,LOO)交互校验(cross-validation,CV)复相关系数(Q_(CV)~2)和外部样本校验复相关系数r~2(test)分别为0.936、0.894、0.772。结果表明,RASMS能较好表征香豆素类衍生物抗艾滋病药物分子的结构信息,且所建模型具有良好稳定性和预测能力,可用于研发新型HIV整合酶抑制剂。     

QSAR/QSPR模型验证方式与预测能力的关系研究

QSAR研究中,判断模型预测能力至关重要。长期以来,模型的预测能力是使用留一法或留k法等内部验证来确定,但在2004年形成的OECD规则中,已明确规定必须使用外部验证集去评价模型的预测能力。为了研究内部验证和外部验证与模型预测能力之间的关系,本文以45种睾酮和二氢睾酮衍生物以及37种萘锟酯衍生物为研究对象,以E-Dragon计算的分子描述符作为自变量,在增n减l算法选择变量的基础上,采用SVM算法对同种物质的不同活性以及不同物质的不同活性建立QSAR模型,研究QSAR/QSPR建模时的不同验证方式与模型预测能力的关系。研究结果表明,模型的预测能力与内部验证结果的好坏无必然联系,而结合外部验证的检验结果则是判断模型预测能力的可靠依据。     

运用线性和非线性的方法预测烷基苯的沸点和摩尔体积

以2D-autocorrelation描述符为结构参数,采用PSO和逐步回归的方法进行变量筛选,再结合SVM等机器学习算法对28种苯丙烯盐类化合物对EBV-EA病毒的抑制性活性进行定量构效关系(QSAR)研究.研究结果表明,PSO-v-SVM模型具有最优的模型稳健性和预测效果.由PSO选入的构成该模型的5个2D-autocorrelation描述符为ATS5v,ATS6e,ATS8e,ATS3p,GATS5p;该模型对训练集的拟合和留一法交叉验证结果的相关系数R~2和q_(cv)~2分别为0.986和0.930,对测试集预测结果的相关系数R~2_(ext)达0.955.对5个变量的理化意义的分析表明,极化率、Van der Waals体积和电负性对苯丙烯盐类化合物的抑制性活性影响分别约占57.13%、15.90%和26.97%.     

脂肪醇化合物对梨形四膜虫急性毒性的QSAR研究

为提高脂肪醇化合物对梨形四膜虫急性毒性的预测精度,提出基于定量结构-活性关系(QSAR)原理的脂肪醇化合物对梨形四膜虫急性毒性预测方法。运用遗传算法筛选出5种分子描述符作为变量,采用多元线性回归方法和最小二乘-支持向量机方法建立基于该5种分子描述符的脂肪醇化合物对梨形四膜虫急性毒性的预测模型。对所建立的模型进行内部验证和外部验证,两种模型的复相关系数、留一法交互验证系数分别为0.984、0.979和0.985、0.982,对外部预测样本的复相关系数和外部测试集交互验证系数分别为0.978、0.977和0.979、0.979。结果表明,所建QSAR模型均具有较好的稳健性、预测能力和泛化性能。LS-SVM模型在精度上略优于ML-R模型,而MLR模型更为简单和方便。     

基于量子化学方法发现新型酪氨酸激酶抑制剂

采用量子化学中的密度泛函理论方法,在B3LYP/6-31++G水平下,系统计算了11种亚苄丙二腈类衍生物的量子化学参数,并通过回归分析方法,构建了二维定量构效模型,分析了影响活性抑制的主要因素,并建立了亚苄丙二腈衍生物与抑制酪氨酸激酶活性之间的定量构效关系方程,研究结果表明:该类化合物分子的分子总能量ET与疏水性参数log P对抑制活性的影响最大,且疏水性能越强,分子的活性抑制能力越高。在此基础上,使用留一法交叉验证了模型的预测能力,结果表明,模型的回归系数和留一法交叉验证系数分别为0.796和0.7291,表明模型具有较好的预测能力,可以用于预测此类化合物的活性。基于结构相似性,设计了4种新型的亚苄丙二腈类衍生物分子,在相同水平下计算其量子化学参数,并预测其活性,结果表明这些新型新型酪氨酸激酶抑制剂均具有较好的活性,研究结果为进一步设计性能更好的酪氨酸激酶抑制剂提供了理论参考。     

Learning multi-linear representations of distributions for efficient inference

We examine the class of multi-linear representations (MLR) for expressing probability distributions over discrete variables. Recently, MLR have been considered as intermediate representations that facilitate inference in distributions represented as graphical models. We show that MLR is an expressive representation of discrete distributions and can be used to concisely represent classes of distributions which have exponential size in other commonly used representations, while supporting probabilistic inference in time linear in the size of the representation. Our key contribution is presenting techniques for learning bounded-size distributions represented using MLR, which support efficient probabilistic inference. We demonstrate experimentally that the MLR representations we learn support accurate and very efficient inference.     

Action recognition by joint learning

Due to the promising applications including video surveillance, video annotation, and interaction gaming, human action recognition from videos has attracted much research interest. Although various works have been proposed for human action recognition, there still exist many challenges such as illumination condition, viewpoint, camera motion and cluttered background. Extracting discriminative representation is one of the main ways to handle these challenges. In this paper, we propose a novel action recognition method that simultaneously learns middle-level representation and classifier by jointly training a multinomial logistic regression (MLR) model and a discriminative dictionary. In the proposed method, sparse code of low-level representation, conducting as latent variables of MLR, can capture the structure of low-level feature and thus is more discriminate. Meanwhile, the training of dictionary and MLR model are integrated into one objective function for considering the information of categories. By optimizing this objective function, we can learn a discriminative dictionary modulated by MLR and a MLR model driven by sparse coding. The proposed method is evaluated on YouTube action dataset and HMDB51 dataset. Experimental results demonstrate that our method is comparable with mainstream methods.     

Predicting the occurrence of adverse events using an adaptive neuro-fuzzy inference system (ANFIS) approach with the help of ANFIS input selection

This study presents an adaptive neuro-fuzzy inference system (ANFIS) approach performed to estimate the number of adverse events where the dependent variables are adverse events leading to four types of variables: number of people killed, wounded, hijacked and total number of adverse events. Fourteen infrastructure development projects were selected based on allocated budgets values at different time periods, population density, and previous month adverse event numbers selected as independent variables. Firstly, number of independent variables was reduced by using ANFIS input selection approach. Then, several ANFIS models were performed and investigated for Afghanistan and the whole country divided into seven regions for analysis purposes. Performances of models were assessed and compared based on the mean absolute errors. The difference between observed and estimated value was also calculated within \({\pm }1\) range with values around 90 %. We included multiple linear regression (MLR) model results to assess the predictive power of the ANFIS approach, in comparison to a traditional statistical approach. When the model accuracy was calculated according to the performance metrics, ANFIS showed greater predictive accuracy than MLR analysis, as indicated by experimental results. As a result of this study, we conclude that ANFIS is able to estimate the occurrence of adverse events according to economical infrastructure development project data.     

Estimating electromyography responses using an adaptive neuro‐fuzzy inference system with subtractive clustering

This study aimed to develop an adaptive neuro‐fuzzy inference system (ANFIS) approach to estimate the normalized electromyography (NEMG) responses, where the independent variables are demographic variables including population, gender, ethnicity, age, height, weight, posture, and muscle groups. The study groups comprised 75 US‐based (54 males and 21 females) and 10 Japan‐based (all males) automobile assembly workers. A total of 65 inputs and 1 output reflecting the NEMG values were considered at the beginning. After correlating analysis results, a total of 35 significant predictors were considered for both ANFIS and regression models. The data were partitioned into two datasets, one for training (70% of all data) and one for validation (30% of all data). In addition to a soft‐computing approach, a multiple linear regression (MLR) analysis was also performed to evaluate whether or not the ANFIS approach showed superior predictive performance compared to a classical statistical approach. According to the performance comparison, ANFIS had better predictive accuracy than MLR, as demonstrated by the experimental results. Overall, this study demonstrates that the ANFIS approach can predict normalized EMG responses according to subjects’ demographic variables, posture, and muscle groups.     

ANFIS and statistical based approach to prediction the peak pressure load of concrete pipes including glass fiber

In this paper, Adaptive Neural Fuzzy Inference System (ANFIS) and Multiple Linear Regression (MLR) models are discussed to determine peak pressure load measurements of the 0, 0.2, 0.4 and 0.6% glass fibers (by weight) reinforced concrete pipes having 200, 300, 400, 500 and 600 mm diameters. For comparing the ANFIS, MLR and experimental results, determination coefficient (R²), root mean square error (RMSE) and standard error of estimates (SEE) statistics were used as evaluation criteria. It is concluded that ANFIS and MLR are practical methods for predicting the peak pressure load (PPL) values of the concrete pipes containing glass fibers and PPL values can be predicted using ANFIS and MLR without attempting any experiments in a quite short period of time with tiny error rates. Furthermore ANFIS model has the predicting potential better than MLR.     

A ‘non-parametric’ version of the naive Bayes classifier

Many algorithms have been proposed for the machine learning task of classification. One of the simplest methods, the naive Bayes classifier, has often been found to give good performance despite the fact that its underlying assumptions (of independence and a normal distribution of the variables) are perhaps violated. In previous work, we applied naive Bayes and other standard algorithms to a breast cancer database from Nottingham City Hospital in which the variables are highly non-normal and found that the algorithm performed well when predicting a class that had been derived from the same data. However, when we then applied naive Bayes to predict an alternative clinical variable, it performed much worse than other techniques. This motivated us to propose an alternative method, based on naive Bayes, which removes the requirement for the variables to be normally distributed, but retains the essential structure and other underlying assumptions of the method. We tested our novel algorithm on our breast cancer data and on three UCI datasets which also exhibited strong violations of normality. We found our algorithm outperformed naive Bayes in all four cases and outperformed multinomial logistic regression (MLR) in two cases. We conclude that our method offers a competitive alternative to MLR and naive Bayes when dealing with data sets in which non-normal distributions are observed.     

Systematic multiscale models to predict the compressive strength of self-compacting concretes modified with nanosilica at different curing ages

The evolution of nanotechnology brings materials with novel performance and during last year’s much attempt has been established to include nanoparticles especially nano-silica (NS) into the concrete to improve performance and develop concrete with enhanced characteristics. Generally, NS is incorporated into the self-compacting concrete (SCC) aiming to positively influence the fresh, mechanical, microstructure, and durability properties of the composite. The most important mechanical property for all types of concrete composites is compressive strength. Therefore, developing reliable models for predicting the compressive strength of SCC is crucial regarding saving time, energy, and cost-effectiveness. Moreover, it gives valuable information for scheduling the construction work and provides information about the correct time for removing the formwork. In this study, three different models including the linear relationship model (LR), nonlinear model (NLR), and multi-logistic model (MLR) were proposed to predict the compressive strength of SCC mixtures made with or without NS. In this regard, a comprehensive data set that consists of 450 samples were collected and analyzed to develop the models. In the modeling process, the most important variables affecting the compressive strength such as NS content, cement content, water to binder ratio, curing time from 1 to 180 days, superplasticizer content, fine aggregate content, and coarse aggregate content were considered as input variables. Various statistical assessments such as Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Scatter Index (SI), OBJ value, and the coefficient of determination (R²) were used to evaluate the performance of the proposed models. The results indicated that the MLR model performed better for forecasting the compression strength of SCC mixtures modified with NS compared to other models. The SI and OBJ values of the MLR model were 18.8% and 16.7% lower than the NLR model, indicating the superior performance of the MLR model. Moreover, the sensitivity analysis demonstrated that the curing time is the most affecting variable for forecasting the compressive strength of SCC modified with NS.

     

Comparison between back propagation neural network and regression models for the estimation of pigment content in rice leaves and panicles using hyperspectral data

Two nitrogen experiments on rice were conducted in 2002, and the reflectances (350 to 2500 nm) and pigment contents (chlorophylls a and b, total chlorophylls and carotenoids) for leaf and panicle samples at different growth stages were measured in the laboratory. After performing an outlier analysis, the number of samples were 843 for leaves and 188 for panicles. Absorption features at 430, 460, 470, 640 and 660 nm for different pigments, and the relative reflectance of the green peak around 550 nm calculated by the continuum‐removed method, as well as the red edge position (REP) of rice leaves and panicles were selected as the independent variables, and measured pigment contents were selected as the dependent variables. Then, back propagation neural network (BPN) models, a kind of artificial neuron network (ANN), and multivariate linear regression models (MLR) were trained and tested. The main objective of this study was to compare the predictive ability of the ANN models to that of the MLR models in estimating the content of pigments in rice leaves and panicles. Results showed that all BPN models gave higher coefficients of determination (R²) and lower absolute errors (ABSEs) and root mean squared errors (RMSEs) than the corresponding MLR models, in both calibration and validation tests. Further significance tests by paired t tests and bootstrapping algorithms indicated that most of the BPN models outperformed the MLR models. When trained by combination data that did not meet the assumption of normal distribution, the BPN models appeared to not only have a better learning ability, but also had a more accurate predictive power than the MLR models. The estimation of leaf pigments was more accurate than that of panicle pigments, independent of which model was used.     

Energy efficiency analysis for flexible-grid OFDM-based optical networks

As the Internet traffic grows, the energy efficiency gains more attention as a design factor for the planning and operation of telecommunication networks. This paper is devoted to the study of energy efficiency in optical transport networks, comparing the performance of an innovative flexible-grid network based on Orthogonal Frequency Division Multiplexing (OFDM) with that of conventional fixed-grid Wavelength Division Multiplexing (WDM) networks with a Single Line Rate (SLR) and with a Mixed Line Rate (MLR) operation. The power consumption values of the network elements are introduced. Energy-aware heuristic algorithms are proposed for the resource allocation both in static (offline) and dynamic (online) scenarios with time-varying demands for the Elastic-bandwidth OFDM-based network and the WDM networks (with SLR and MLR). The energy efficiency performance of the two network technologies under different traffic load conditions have been demonstrated for different network sizes through simulations based on the proposed algorithms. The results in energy efficiency and network blocking highlight the benefits of the bandwidth elasticity and the flexibility of selecting different modulation formats offered by OFDM networks.