Andrew M. Woodward

Rapid identification of urinary tract infection bacteria using hyperspectral whole-organism fingerprinting and artificial neural networks

Three rapid spectroscopic approaches for whole-organism fingerprinting-pyrolysis mass spectrometry (PyMS), Fourier transform infra-red spectroscopy (FT-IR) and dispersive Raman microscopy--were used to analyse a group of 59 clinical bacterial isolates associated with urinary tract infection. Direct visual analysis of these spectra was not possible, highlighting the need to use methods to reduce the dimensionality of these hyperspectral data. The unsupervised methods of discriminant function and hierarchical cluster analyses were employed to group these organisms based on their spectral fingerprints, but none produced wholly satisfactory groupings which were characteristic for each of the five bacterial types. In contrast, for PyMS and FT-IR, the artificial neural network (ANN) approaches exploiting multi-layer perceptrons or radial basis functions could be trained with representative spectra of the five bacterial groups so that isolates from clinical bacteriuria in an independent unseen test set could be correctly identified. Comparable ANNs trained with Raman spectra correctly identified some 80% of the same test set. PyMS and FT-IR have often been exploited within microbial systematics, but these are believed to be the first published data showing the ability of dispersive Raman microscopy to discriminate clinically significant intact bacterial species. These results demonstrate that modern analytical spectroscopies of high intrinsic dimensionality can provide rapid accurate microbial characterization techniques, but only when combined with appropriate chemometrics.

An introduction to wavelet transforms for chemometricians: A time-frequency approach

Abstract One way to obtain an intuitive understanding of the wavelet transform is to explain it in terms of segmentation of the time-frequency/scale domain. The ordinary Fourier transform does not contain information about frequency changes over time and the short time Fourier transform (STFT) technique was suggested as a solution to this problem. The wavelet transform has similarities to STFT, but partitions the time-frequency space differently in order to obtain better resolutions along time and frequency/scales. In STFT a constant bandwidth partitioning is performed whereas in the wavelet transform the time-frequency domain is partitioned according to a constant relative bandwidth scheme. In this paper we also discuss the following application areas of wavelet transforms in chemistry and analytical biotechnology: denoising, removal of baselines, determination of zero crossings of higher derivatives, signal compression and wavelet preprocessing in partial least squares (PLS) regression.

Wavelet Denoising of Infrared Spectra

The application of wavelet denoising to infrared spectra was investigated. Six different wavelet denoising methods (SURE, VISU, HYBRID, MINMAX, MAD and WAVELET PACKETS) were applied to pure infrared spectra with various added levels of homo- and heteroscedastic noise. The performances of the wavelet denoising methods were compared with the standard Fourier and moving mean filtering in terms of root mean square errors between the pure and denoised spectra and visual quality of the denoised spectrum. The use of predictive ability as a possible objective criterion for denoising performance was also investigated. The main conclusion is that for very low signal-to-noise ratios (S/N) the standard denoising methods (Fourier and moving mean) are comparable to the more sophisticated methods. At higher S/N levels the wavelet denoising methods, in particular the HYBRID and VISU methods, are better. Wavelet methods are also better in restoring the visual quality of the denoised infrared spectra.

Variable selection in wavelet regression models

Variable selection and compression are often used to produce more parsimonious regression models. But when they are applied directly to the original spectrum domain, it is not easy to determine the type of feature the selected variables represent. By performing variable selection in the wavelet domain we show that it is possible to identify important variables as being part of short- or large-scale features. Therefore, the suggested method is to extract information about the selected variables that otherwise would have been inaccessible. We are also able to obtain information about the location of these features in the original domain. In this article we demonstrate three types of variable selection methods applied to the wavelet domain: selection of optimal combination of scales, thresholding based on mutual information and truncation of weight vectors in the partial least squares (PLS) regression algorithm. We found that truncation of weight vectors in PLS was the most effective method for selecting variables. For the two experimental data sets tested we obtained approximately the same prediction error using less than 1% (for Data set 1) and 10% (for Data set 2) of the original variables. We also discovered that the selected variables were restricted to a limited number of wavelet scales. This information can be used to suggest whether the underlying features may be dominated by narrow (selective) peaks (indicated by variables in short wavelet scale regions) or by broader regions (indicated by variables in long wavelet scale regions). Thus, wavelet regression is here used as an extension of the more traditional Fourier regression (where the modelling is performed in the frequency domain without taking into consideration any of the information in the time domain).

Diffuse reflectance absorbance spectroscopy taking in chemometrics (DRASTIC). A hyperspectral FT-IR-based approach to rapid screening for metabolite overproduction

We introduce diffuse-reflectance absorbance spectroscopy in the mid-infrared as a novel method of chemical imaging for the rapid screening of biological samples for metabolite overproduction, using mixtures of ampicillin with Escherichia coli and Staphylococcus aureus as model systems. Deconvolution of the hyperspectral information provided by the raw diffuse reflectance-absorbance mid-infrared spectra was achieved using a combination of principal components analysis (PCA), artificial neural networks (ANNs) and partial least squares regression (PLS). Whereas a univariate approach necessitates appropriate data selection to remove any interferences, the chemometrics/hyperspectral approach could be employed to permit filtering of undesired components to give accurate quantification by PLS and ANNs without any preprocessing. The use of PCs as inputs to the ANNs decreased the training time from some 12 h to ca. 5 min. Equivalent concentrations of ampicillin between 0.05 and 20 mM in an E. coli or S. aureus background were quantified with >95% accuracy using this approach.

Noninvasive, On-Line Monitoring of the Biotransformation by Yeast of Glucose to Ethanol Using Dispersive Raman Spectroscopy and Chemometrics

We describe the first application of dispersive Raman spectroscopy using a diode laser exciting at 780 nm and a charge-coupled device (CCD) detector to the noninvasive, on-line determination of the biotransformation by yeast of glucose to ethanol. Software was developed which automatically removed the effects of cosmic rays and other noise, normalized the spectra to an invariant peak, then removed the “baseline” arising from interference by fluorescent impurities, to obtain the “true” Raman spectra. Variable selection was automatically performed on the parameters of relevant Raman peaks (height, width, position of top and center, area and skewness), and a small subset used as the input to cross-validated models based on partial least-squares (PLS) regression. The multivariate calibration models so formed were sufficiently robust to be able to predict the concentration of glucose and ethanol in a completely different fermentation with a precision better than 5%. Dispersive Raman spectroscopy, when coupled with the appropriate chemometrics, is a very useful approach to the noninvasive, on-line determination of the progress of microbial fermentations.

Genetic programming as an analytical tool for non-linear dielectric spectroscopy

By modelling the non-linear effects of membranous enzymes on an applied oscillating electromagnetic field using supervised multivariate analysis methods, Non-Linear Dielectric Spectroscopy (NLDS) has previously been shown to produce quantitative information that is indicative of the metabolic state of various organisms. The use of Genetic Programming (GP) for the multivariate analysis of NLDS data recorded from yeast fermentations is discussed, and GPs are compared with previous results using Partial Least Squares (PLS) and Artificial Neural Nets (NN). GP considerably outperforms these methods, both in terms of the precision of the predictions and their interpretability.

Dual-frequency excitation: A novel method for probing the nonlinear dielectric properties of biological systems, and its application to suspensions of S. cerevisiae

We have recently described the construction of a dual-cell, nonlinear dielectric spectrometer, and its application to the study of cell suspensions of S. cereuisiae (A.M. Woodward and D.B. Kell, Bioelectrothem. Bioenerg., 24 (1990) 83). Substantial, odd harmonics were generated by these cells when stimulated by very modest sinusoidal electrical fields, within fairly sharp voltageand frequency windows (ca. 0.8-2.5 V cm-‘, l-50 Hz). Resting cells were found to generate only odd-numbered harmonics. In the present work, we have simultaneously applied two sinusoidal frequencies which were indiuidually of unsuitable frequency and/or amplitude for the generation of harmonics when applied to suspensions of S. cerevisiae. Strong “sidebands” or “beat frequencies” were observed which were the (odd-numbered) sums and differences of the exciting frequencies (viz. fi f 2f2, f2 +2f,). The generation of these beat frequencies was strongly inhibited by low concentrations of sodium metavanadate, suggesting that they may be ascribed largely to the H+-ATPase present in the plasma membranes of these cells. We show that the ability of dc fields to inhibit the manifestation of nonlinear dielectric behaviour by these cells is explicable in terms of their ability to act as a field of zero Hz, forcing the excitation out of the amplitude window. When the cells were allowed to glycolyse, beat frequencies of even order ( fi f f*, fi f 3fi) were observed. The present approach provides a novel and powerful approach to the registration of nonlinear dielectric spectra, which, due to the greater precision inherent in the discrimination of frequencies rather than voltages may be expected to provide a more sensitive means of detecting nonlinear dielectric properties in biological systems. If the transduction of exogenous electrical field energy recorded by this method is representative of the natural turnover of the H+-ATPase in viva, it may be calculated that the efficiency of the capture of electric field energy by this enzyme is some 3%.

Rapid Analysis of High-Dimensional Bioprocesses Using Multivariate Spectroscopies and Advanced Chemometrics

There are an increasing number of instrumental methods for obtaining data from biochemical processes, many of which now provide information on many (indeed many hundreds) of variables simultaneously. The wealth of data that these methods provide, however, is useless without the means to extract the required information. As instruments advance, and the quantity of data produced increases, the fields of bioinformatics and chemometrics have consequently grown greatly in importance. The chemometric methods nowadays available are both powerful and dangerous, and there are many issues to be considered when using statistical analyses on data for which there are numerous measurements (which often exceed the number of samples). It is not difficult to carry out statistical analysis on multivariate data in such a way that the results appear much more impressive than they really are. The authors present some of the methods that we have developed and exploited in Aberystwyth for gathering highly multivariate data from bioprocesses, and some techniques of sound multivariate statistical analyses (and of related methods based on neural and evolutionary computing) which can ensure that the results will stand up to the most rigorous scrutiny.

Rapid and non-invasive quantification of metabolic substrates in biological cell suspensions using non-linear dielectric spectroscopy with multivariate calibration and artificial neural networks. Principles and applications

Abstract By studying the non-linear effects of membranous enzymes on an applied oscillating electromagnetic field, non-linear dielectric spectroscopy has previously been shown to produce qualitative information which is indicative of the metabolic state of a variety of organisms. In this study, we extend the method of non-linear dielectric spectroscopy to the production of data sets suitable for use with supervised multivariate analysis methods, in order to allow quantitative prediction of analyte concentrations in unknown samples, again using the alteration in the non-linear dielectric profile produced by these analytes via the metabolism of the cell (as effected via the operation of their membranous enzymes). Non-stationarity in the extent of non-linear electrode polarization can interfere with the measurement of non-linear dielectric spectra; various electrode materials and configurations have been tested for their suitability for use in non-linear dielectric spectroscopy. We exploit partial least-squares regression and artificial neural networks for the multivariate analysis of non-linear dielectric data recorded from yeast cell suspensions, and schemes for preprocessing these data to improve the precision of the prediction of analyte levels are developed and optimized. The resulting analytical methods are applied to the prediction of glucose levels in sheep and human blood, by both invasive and non-invasive measurements, and to the non-invasive measurement of process variables during a microbial fermentation.