Kurt Varmuza

On Entropy-Based Molecular Descriptors: Statistical Analysis of Real and Synthetic Chemical Structures

This paper presents an analysis of entropy-based molecular descriptors. Specifically, we use real chemical structures, as well as synthetic isomeric structures, and investigate properties of and among descriptors with respect to the used data set by a statistical analysis. Our numerical results provide evidence that synthetic chemical structures are notably different to real chemical structures and, hence, should not be used to investigate molecular descriptors. Instead, an analysis based on real chemical structures is favorable. Further, we find strong hints that molecular descriptors can be partitioned into distinct classes capturing complementary information.

Feature selection by genetic algorithms for mass spectral classifiers

Mass spectral classifiers for 15 substructures have been computed that give discrete present/absent answers. For the development of classifiers, linear discriminant analysis (LDA) and partial least squares discriminant PLS (DPLS) have been used. The low resolution mass spectra were transformed into a set of 400 spectral features. Because each spectrum is described with so many features, some features may not be necessary, and others may contribute only noise. Therefore, the effect of feature selection has been investigated. The methods used were selection by Fisher ratios and selection by a genetic algorithm (GA). The first method is univariate, the second is multivariate; advantages and disadvantages of both are discussed. On the average, feature selection did not significantly change the classification performance compared with results that have been obtained with all features. However, it was possible to reduce the number of features considerably without a loss of classification performance. For a few substructures GA together with LDA resulted in much better classifiers than DPLS with all features. The features selected for classifications of a benzyl substructure and for the presence of chlorine have been interpreted in terms of mass spectrometric fragmentation rules.

Spectral similarity versus structural similarity: infrared spectroscopy

A new method is described for evaluation of spectral similarity searches. Aim of the method is to measure the similarity between the chemical structures of query compounds and the found reference compounds (hits). A high structural similarity is essential if the query is not present in the spectral library. Similarity of chemical structures was measured by the Tanimoto index, calculated from 1365 binary substructure descriptors. The method has been applied to several 1000 hitlists from searches in an infrared (IR) spectra database containing 13,484 compounds. Hitlists with highest structure information were obtained using a similarity measure based on the correlation coefficient computed from mean centered absorbance units. Frequency distributions of spectral and structural similarities have been investigated and a threshold for the spectral similarity has been derived that in general gives hitlists exhibiting significant chemical structure similarities with the query.

Information Indices with High Discriminative Power for Graphs

In this paper, we evaluate the uniqueness of several information-theoretic measures for graphs based on so-called information functionals and compare the results with other information indices and non-information-theoretic measures such as the well-known Balaban index. We show that, by employing an information functional based on degree-degree associations, the resulting information index outperforms the Balaban index tremendously. These results have been obtained by using nearly 12 million exhaustively generated, non-isomorphic and unweighted graphs. Also, we obtain deeper insights on these and other topological descriptors when exploring their uniqueness by using exhaustively generated sets of alkane trees representing connected and acyclic graphs in which the degree of a vertex is at most four.

Classification of mass spectra. A comparison of yes/no classification methods for the recognition of simple structural properties

Abstract Mass spectral classifiers for 54 structural properties have been developed using a library of 90 000 spectra and applying four complementary classification methods ( k -nearest neighbor, linear discriminant analysis, SIMCA, and a neural network). Neural networks yielded best results; however, only a few structural properties could be classified with a sufficiently high predictive ability. The main difficulties to be solved for future work are the construction of a library with high-quality spectra, and the definition of structural properties that are actually reflected in low resolution mass spectra. Werther, W., Lohninger, H., Stancl, F. and Varmuza, K., 1994. Classification of mass spectra. A comparison of yes/no classification methods for the recognition of simple structural properties. Chemometrics and Intelligent Laboratory Systems , 22: 63–76.

Determination of glucose and ethanol in bioethanol production by near infrared spectroscopy and chemometrics.

The concentrations of glucose and ethanol in substrates from bioethanol processes have been modeled by near infrared (NIR) spectroscopy data. NIR spectra were acquired in the wavelength range of 1100-2300 nm by means of a transflectance probe for measurements in liquid samples. For building of regression models a genetic algorithm has been applied for variable selection, and partial least-squares (PLS) regression for creation of linear models. A realistic estimation of the prediction performance of the models was obtained by a repeated double cross-validation (rdCV). Reduced data sets with only 15 variables showed improved prediction qualities, in comparison with models containing 235 variables, particularly for the determination of the ethanol concentration in distillation residues (stillages). The squared correlation coefficient, R(2), between the concentrations obtained by HPLC analysis and the concentrations derived from NIR data (using 15 selected wavelengths, test set samples) was 0.999 for ethanol in stillage, and 0.977 for glucose in mash. The standard deviation of prediction errors, SEP, obtained from test set samples was 0.6 g L(-1) for ethanol (2% of the mean ethanol concentration), and 2.0 g L(-1) for glucose (9.6% of the mean glucose concentration).

Comet 67P/Churyumov–Gerasimenko: Close-up on Dust Particle Fragments

The COmetary Secondary Ion Mass Analyser instrument on board ESAs Rosetta mission has collected dust particles in the coma of comet 67P/Churyumov-Gerasimenko. During the early-orbit phase of the Rosetta mission, particles and particle agglomerates have been imaged and analyzed in the inner coma at distances between 100 km and 10 km off the cometary nucleus and at more than 3 AU from the Sun. We identified 585 particles of more than 14 μm in size. The particles are collected at low impact speeds and constitute a sample of the dust particles in the inner coma impacting and fragmenting on the targets. The sizes of the particles range from 14 μm up to sub-millimeter sizes and the differential dust flux size distribution is fitted with a power law exponent of -3.1. After impact, the larger particles tend to stick together, spread out or consist of single or a group of clumps, and the flocculent morphology of the fragmented particles is revealed. The elemental composition of the dust particles is heterogeneous and the particles could contain typical silicates like olivine and pyroxenes, as well as iron sulfides. The sodium to iron elemental ratio is enriched with regard to abundances in CI carbonaceous chondrites by a factor from ˜1.5 to ˜15. No clear evidence for organic matter has been identified. The composition and morphology of the collected dust particles appear to be similar to that of interplanetary dust particles.

High-molecular-weight organic matter in the particles of comet 67P/Churyumov–Gerasimenko

The presence of solid carbonaceous matter in cometary dust was established by the detection of elements such as carbon, hydrogen, oxygen and nitrogen in particles from comet 1P/Halley. Such matter is generally thought to have originated in the interstellar medium, but it might have formed in the solar nebula—the cloud of gas and dust that was left over after the Sun formed. This solid carbonaceous material cannot be observed from Earth, so it has eluded unambiguous characterization. Many gaseous organic molecules, however, have been observed; they come mostly from the sublimation of ices at the surface or in the subsurface of cometary nuclei. These ices could have been formed from material inherited from the interstellar medium that suffered little processing in the solar nebula. Here we report the in situ detection of solid organic matter in the dust particles emitted by comet 67P/Churyumov–Gerasimenko; the carbon in this organic material is bound in very large macromolecular compounds, analogous to the insoluble organic matter found in the carbonaceous chondrite meteorites. The organic matter in meteorites might have formed in the interstellar medium and/or the solar nebula, but was almost certainly modified in the meteorites’ parent bodies. We conclude that the observed cometary carbonaceous solid matter could have the same origin as the meteoritic insoluble organic matter, but suffered less modification before and/or after being incorporated into the comet.

Severe drought stress is affecting selected primary metabolites, polyphenols, and volatile metabolites in grapevine leaves (Vitis vinifera cv. Pinot noir)

Extreme weather conditions with prolonged dry periods and high temperatures as well as heavy rain events can severely influence grapevine physiology and grape quality. The present study evaluates the effects of severe drought stress on selected primary metabolites, polyphenols and volatile metabolites in grapevine leaves. Among the 11 primary metabolites, 13 polyphenols and 95 volatiles which were analyzed, a significant discrimination between control and stressed plants of 7 primary metabolites, 11 polyphenols and 46 volatile metabolites was observed. As single parameters are usually not specific enough for the discrimination of control and stressed plants, an unsupervised (PCA) and a supervised (PLS-DA) multivariate approach were applied to combine results from different metabolic groups. In a first step a selection of five metabolites, namely citric acid, glyceric acid, ribose, phenylacetaldehyde and 2-methylbutanal were used to establish a calibration model using PLS regression to predict the leaf water potential. The model was strong enough to assign a high number of plants correctly with a correlation of 0.83. The PLS-DA provides an interesting approach to combine data sets and to provide tools for the specific evaluation of physiological plant stresses.

A Large Scale Analysis of Information-Theoretic Network Complexity Measures Using Chemical Structures

This paper aims to investigate information-theoretic network complexity measures which have already been intensely used in mathematical- and medicinal chemistry including drug design. Numerous such measures have been developed so far but many of them lack a meaningful interpretation, e.g., we want to examine which kind of structural information they detect. Therefore, our main contribution is to shed light on the relatedness between some selected information measures for graphs by performing a large scale analysis using chemical networks. Starting from several sets containing real and synthetic chemical structures represented by graphs, we study the relatedness between a classical (partition-based) complexity measure called the topological information content of a graph and some others inferred by a different paradigm leading to partition-independent measures. Moreover, we evaluate the uniqueness of network complexity measures numerically. Generally, a high uniqueness is an important and desirable property when designing novel topological descriptors having the potential to be applied to large chemical databases.