The application of artificial neural networks for the selection of key thermoanalytical parameters in medicinal plants analysis

In the present study three thermoanalytical methods: differential thermal analysis (DTA), thermogravimetric analysis (TGA), and derivative thermogravimetric analysis (DTG) were used to investigate the thermal behavior of medicinal plant raw materials. In order to describe DTA curve, designation of the onset T(i), and peak T(p), temperatures was required. In TGA the mass losses Delta(m), and in DTG the temperature range of peak DeltaT, peak temperature T(p), and peak height h, were recorded. All parameters were read for three stages of the thermal decomposition of plant samples which resulted in obtaining eighteen thermal variables for each sample. Some similarities in the course of thermal decomposition of the same plant organs were recognized, but complexity of the obtained data made it very difficult to determine if they could differentiate between medicinal plant materials and which of them encode the most valuable information about the studied herbals. In order to confirm the existence of any relations between the chemical composition of medicinal plants and their thermal decomposition and to find out which thermoanalytical variables or decomposition stages can be considered as the most significant in terms of their evaluation, it was decided to apply fully connected feed-forward artificial neural networks (ANN). Two different training algorithms were used to address the problem: back-propagation of error and conjugate gradient descent. To verify the results two-dimensional (2-D) Kohonen self-organizing feature maps (SOFMs) were employed. Two alternative datasets of thirteen key variables discriminating plant samples have been proposed.     

The use of artificial neural networks for the selection of the most appropriate thermal parameters and for the classification of a set of phenylcarbamic acid derivates

The objective of this work was to apply artificial neural networks (ANNs) to the classification group of 43 derivatives of phenylcarbamic acid. To find the appropriate clusters Kohonen topological maps were employed. As input data, thermal parameters obtained during DSC and TG analysis were used. Input feature selection (IFS) algorithms were used in order to give an estimate of the relative importance of various input variables. Additionally, sensitivity analysis was carried out to eliminate less important thermal variables. As a result, one classification model was obtained, which can assign our compounds to an appropriate class. Because the classes contain groups of molecules structurally related, it is possible to predict the structure of the compounds (for example the position of the substitution alkoxy group in the phenyl ring) on the basis of obtained parameters.     

Characterization of Galician (N.W. Spain) quality brand potatoes: a comparison study of several pattern recognition techniques

Authenticity is an important food quality criterion and rapid methods to guarantee it are widely demanded by food producers, processors, consumers and regulatory bodies. The objective of this work was to develop a classification system in order to confirm the authenticity of Galician potatoes with a Certified Brand of Origin and Quality (CBOQ) 'Denominación Específica: Patata de Galicia' and to differentiate them from other potatoes that did not have this CBOQ. Ten selected metals were determined by atomic spectroscopy in 102 potato samples which were divided into two categories: CBOQ and non-CBOQ potatoes. Multivariate chemometric techniques, such as cluster analysis and principal component analysis, were applied to perform a preliminary study of the data structure. Four supervised pattern recognition procedures [including linear discriminant analysis (LDA), K-nearest neighbours (KNN), soft independent modelling of class analogy (SIMCA) and multilayer feed-forward neural networks (MLF-ANN)] were used to classify samples into the two categories considered on the basis of the chemical data. Results for LDA, KNN and MLF-ANN are acceptable for the non-CBOQ class, whereas SIMCA showed better recognition and prediction abilities for the CBOQ class. A more sophisticated neural network approach performed by the combination of the self-organizing with adaptive neighbourhood network (SOAN) and MLF network was employed to optimize the classification. Using this combined method, excellent performance in terms of classification and prediction abilities was obtained for the two categories with a success rate ranging from 98 to 100%. The metal profiles provided sufficient information to enable classification rules to be developed for identifying potatoes according to their origin brand based on SOAN-MLF neural networks.     

Chemometrics characterisation of the quality of river water

Within the period from autumn 1990 to spring 1999 (from October to April in each period) 207 samples were collected and the measurement of 19 physical and chemical variables of the Mura river, Slovenia, were carried out. These variables are: river flow, water temperature, air temperature, dissolved oxygen, deficit of oxygen, oxygen saturation index, chemical oxygen demand (COD) in unfiltered and filtered samples, and biochemical oxygen demand after 5 days (BOD5) in unfiltered and filtered samples, pH, conductivity, ammonium, nitrite, nitrate, and phosphate concentrations, adsorbable organic halogens (AOX), dissolved organic carbon (DOC), and suspended solids. For handling the results of all measurements different chemometrics methods were employed: (i) the basic statistical methods for the determination of mean and median values, standard deviations, minimal and maximal values of measured variables, and their mutual correlation coefficients, (ii) the principal component analysis (PCA), and (iii) the clustering method based on Kohonen neural network. The influences of season, month, sampling site, and sampling time on the pollutant levels were examined. Before 1993, the pulp and paper industry was the main source of pollutants because of large amounts of chlorine emission as a consequence of industrial treatment, the leaching of cellulose. After the year 1993, the technology was changed and the quality of the river water has improved. The improvement could be detected 1 year after the change of technology. For one part of water samples the river quality classes based on biological parameters were also determined. The correlation between the biologically determined quality classes and chemical measurements was sought. Consequently, the biological classification for the water samples based on the chemical analyses was studied.     

Neural Networks as a Tool for Quality Assessment of Soybean and Rapeseed Oils Analyzed by Thermogravimetry

The quality of commercial vegetable oils is usually evaluated via chemical parameters such as density, refractive index, saponification, iodine and acid values. In this paper, the applicability of thermal parameters for the quality assessment of vegetable oils is proposed. In order to achieve this goal, different back-propagation neural network architectures were trained, using chemical and thermal parameters as inputs. To avoid any accidental correlation due to the random initialization of the weights, each topology was repeated three times and three networks were chosen, with 5-3-2, 8-5-2 and 13-6-2 structures. The error function sum square error (SSE) was used as the criterion for finalization of the learning process. A model was developed for the correct classification of oils with regard to their type and quality. This revised version was published online in August 2006 with corrections to the Cover Date.     

On the use of counterpropagation artificial neural networks to characterize Italian rice varieties

A counterpropagation artificial neural network (CP-ANN) approach was used to classify 1779 Italian rice samples according to their variety, using physical measurements which are routinely determined for the commercial classification of the product. If compared to the classical Principal Component Analysis, the mapping based on the Kohonen network showed a significantly better representational ability, being able to separate classes which appeared quite undistinguished in the PC space. From the classification and prediction viewpoint, the optimal CP-ANN was able to correctly predict more than 90% of the test set samples.     

The application of Kohonen neural networks to diagnose calibration problems in atomic absorption spectrometry

In atomic absorption spectrometric measurements calibration lines are measured daily. These lines are not always acceptable. They can, for instance, contain outliers, have a bad precision or can be curved. To evaluate the quality of those lines a method which gives a fast diagnosis is recommended. In this study the use of Kohonen neural networks was examined as an automated procedure to classify these calibration lines. The results were compared with those obtained using a decision support system which uses classical statistical methods to classify the lines. The prediction capabilities of both approaches relative to a visual inspection and classification was found to be comparable, or even slightly better for the Kohonen networks, depending on the training set used. For both techniques a prediction error rate of <10% was obtained, relative to a visual classification.     

Application of high-order neural networks in chemistry

Summary The effectiveness and usefulness of the so-called high-order neural networks for classification of chemical objects is demonstrated. The high-order neural networks usually do not need hidden neurons for correct interpretation of patterns. A simple formula for partial derivatives of the minimized objective (error) function is derived, which is used for production of weight coefficients during the adaptation process. An illustrative example dealing with inductive and resonance effects of functional groups by the second-order neural network is presented.     

Application of artificial neural networks to the classification of soils from São Paulo State using near-infrared spectroscopy

This paper describes how artificial neural networks can be used to classify multivariate data. Two types of neural networks were applied: a counter propagation neural network (CP-ANN) and a radial basis function network (RBFN). These strategies were used to classify soil samples from different geographical regions in Brazil by means of their near-infrared (diffuse reflectance) spectra. The results were better with CP-ANN (classification error 8.6%) than with RBFN (classification error 11.0%).     

Validation using sensitivity and target transform factor analyses of neural network models for classifying bacteria from mass spectra

Temperature constrained cascade correlation networks (TCCCNs) are computational neural networks that configure their own architecture, train rapidly, and give reproducible prediction results. TCCCN classification models were built using the Latin-partition method for five classes of pathogenic bacteria. Neural networks are problematic in that the relationships among the inputs (i.e., mass spectra) and the outputs (i.e., the bacterial identities) are not apparent. In this study, neural network models were constructed that successfully classified the targeted bacteria and the classification model was validated using sensitivity and target transformation factor analysis (TTFA). Without validation of the classification model, it is impossible to ascertain whether the bacteria are classified by peaks in the mass spectrum that have no causal relationships with the bacteria, but instead randomly correlate with the bacterial classes. Multiple single output network models did not offer any benefits when compared to single network models that had multiple outputs. A multiple output TCCCN model achieved classification accuracies of 96 +/- 2% and exhibited improved performance over multiple single output TCCCN models. Chemical ionization mass spectra were obtained from in situ thermal hydrolysis methylation of freeze-dried bacteria. Mass spectral peaks that pertain to the neural network classification model of the pathogenic bacterial classes were obtained by sensitivity analysis. A significant number of mass spectral peaks that had high sensitivity corresponded to known biomarkers, which is the first time that the significant peaks used by a neural network model to classify mass spectra have been divulged. Furthermore, TTFA furnishes a useful visual target as to which peaks in the mass spectrum correlate with the bacterial identities.     

人工神经网络在溶解度预测方面的应用

溶解度作为一项重要的物化指标,一直是化学学科的研究重点。然而,通过实验测量获得数据耗时费力,因此,科研人员建立了多种理论方法来进行估算,其中,人工神经网络因其能够关联复杂的多变量情况而受到广泛关注。本文综述了人工神经网络在物质溶解度预测方面的应用,介绍了应用最广泛的3种神经网络(BP神经网络、小波神经网络、径向基神经网络)的模型结构、预测方法和预测优势,探讨了神经网络的不足以及改进方法。文章最后对神经网络在物质溶解度预测方面的发展前景进行了展望。与其他方法相比,人工神经网络技术在物质溶解度预测方面具有预测结果精确度高、操作简单等特点,具有广阔的应用前景,但输入变量选择、隐含层节点数确定、避免局部最优等问题还需逐步建立系统的理论指导。     

Evaluation of chemometric techniques and artificial neural networks for cancer screening using Cu, Fe, Se and Zn concentrations in blood serum

It is known that variations in the concentrations of certain trace elements in bodily fluids may be an indication of an alteration of the organism's normal functioning. Multivariate analysis of such data, and its comparison against proper reference values, can thus be employed as possible screening tests. Such analysis has usually been conducted by means of chemometric techniques and, to a lower extent, backpropagation artificial neural networks. Despite the excellent classification capacities of the latter, its development can be time-consuming and computer-intensive. Probabilistic artificial neural networks represent another attractive, yet scarcely evaluated classification technique which could be employed for the same purpose. The present work was aimed at comparing the performance of two chemometric techniques (principal component analysis and logistic regression) and two artificial neural networks (a backpropagation and a probabilistic neural network) as screening tools for cancer. The concentrations of copper, iron, selenium and zinc, which are known to play an important role in body chemistry, were used as indicators of physical status. Such elements were determined by total reflection X-ray fluorescence spectrometry in a sample of blood serum taken from individuals who had been given a positive (N = 27) or a negative (N = 32) diagnostic on various forms of cancer. The principal components analysis served two purposes: (i) initial screening of the data; and, (ii) reducing the dimension of the data space to be input to the networks. The logistic regression, as well as both artificial neural networks showed an outstanding performance in terms of distinguishing healthy (specificity: 90-100%) or unhealthy individuals (sensitivity: 100%). The probabilistic neural network offered additional advantages when compared to the more popular backpropagation counterpart. Effectively, the former is easier and faster to develop, for a smaller number of variables has to be optimized, and there are no risk of overtraining.     

Comparing the performance of neural networks to well-established methods of multivariate data analysis: the classification of mass spectral data

Summary The performance of neural networks in classifying mass spectral data is evaluated and compared to methods of multivariate data analysis and pattern recognition. Back propagation networks are matched with linear discriminant analysis, Kohonen feature maps are compared to the knearest neighbour clustering algorithm. Eight classifiers were trained, in order to discriminate mass spectra of steroids from eight distinct classes of chemical compounds. The results obtained show slightly better performance of Kohonen networks compared to k-nearest neighbour clustering and equal performance of multi-layer perceptrons and discriminant analysis.     

Predicting physical-chemical properties of compounds from molecular structures by recursive neural networks

In this paper, we report on the potential of a recently developed neural network for structures applied to the prediction of physical chemical properties of compounds. The proposed recursive neural network (RecNN) model is able to directly take as input a structured representation of the molecule and to model a direct and adaptive relationship between the molecular structure and target property. Therefore, it combines in a learning system the flexibility and general advantages of a neural network model with the representational power of a structured domain. As a result, a completely new approach to quantitative structure-activity relationship/quantitative structure-property relationship (QSPR/QSAR) analysis is obtained. An original representation of the molecular structures has been developed accounting for both the occurrence of specific atoms/groups and the topological relationships among them. Gibbs free energy of solvation in water, Delta(solv)G degrees , has been chosen as a benchmark for the model. The different approaches proposed in the literature for the prediction of this property have been reconsidered from a general perspective. The advantages of RecNN as a suitable tool for the automatization of fundamental parts of the QSPR/QSAR analysis have been highlighted. The RecNN model has been applied to the analysis of the Delta(solv)G degrees in water of 138 monofunctional acyclic organic compounds and tested on an external data set of 33 compounds. As a result of the statistical analysis, we obtained, for the predictive accuracy estimated on the test set, correlation coefficient R = 0.9985, standard deviation S = 0.68 kJ mol(-1), and mean absolute error MAE = 0.46 kJ mol(-1). The inherent ability of RecNN to abstract chemical knowledge through the adaptive learning process has been investigated by principal components analysis of the internal representations computed by the network. It has been found that the model recognizes the chemical compounds on the basis of a nontrivial combination of their chemical structure and target property.     

基于独立分量和神经网络的近红外多组分分析方法

采用小波变换对光谱数据进行压缩,用独立分量分析(ICA)方法提取近红外光谱数据矩阵的独立成分和相应的混合矩阵,再用BP神经网络对混合矩阵和实测浓度矩阵进行建模,提出了基于独立分量分析-神经网络回归(ICA-NNR)的近红外分析建模方法。进一步研究了独立分量数和网络中间隐层的神经元数对模型性能的影响,经优化后的ICA-NNR模型在相关系数与均方根误差两个指标上均优于直接用光谱矩阵作为输入所建立的模型。本方法用于玉米中水分、淀粉、蛋白质3种主要成分含量的同时测定,检验样品集的化学检测值与近红外预测值的相关系数分别达到:淀粉r=0.971,蛋白质r=0.976,水分r=0.975。     

Validation of counter propagation neural network models for predictive toxicology according to the OECD principles: a case study

The OECD has proposed five principles for validation of QSAR models used for regulatory purposes. Here we present a case study investigating how these principles can be applied to models based on Kohonen and counter propagation neural networks. The study is based on a counter propagation network model that has been built using toxicity data in fish fathead minnow for 541 compounds. The study demonstrates that most, if not all, of the OECD criteria may be met when modeling using this neural network approach.