Richard J. Gilbert

Explanatory Analysis of the Metabolome Using Genetic Programming of Simple, Interpretable Rules

Genetic programming, in conjunction with advanced analytical instruments, is a novel tool for the investigation of complex biological systems at the whole-tissue level. In this study, samples from tomato fruit grown hydroponically under both high- and low-salt conditions were analysed using Fourier-transform infrared spectroscopy (FTIR), with the aim of identifying spectral and biochemical features linked to salinity in the growth environment. FTIR spectra of whole tissue extracts are not amenable to direct visual analysis, so numerical modelling methods were used to generate models capable of classifying the samples based on their spectral characteristics. Genetic programming (GP) provided models with a better prediction accuracy to the conventional data modelling methods used, whilst being much easier to interpret in terms of the variables used. Examination of the GP-derived models showed that there were a small number of spectral regions that were consistently being used. In particular, the spectral region containing absorbances potentially due to a cyanide/nitrile functional group was identified as discriminatory. The explanatory power of the GP models enabled a chemical interpretation of the biochemical differences to be proposed. The combination of FTIR and GP is therefore a powerful and novel analytical tool that, in this study, improves our understanding of the biochemistry of salt tolerance in tomato plants.

Genetic Programming: A Novel Method for the Quantitative Analysis of Pyrolysis Mass Spectral Data

A technique for the analysis of multivariate data by genetic programming (GP) is described, with particular reference to the quantitative analysis of orange juice adulteration data collected by pyrolysis mass spectrometry (PyMS). The dimensionality of the input space was reduced by ranking variables according to product moment correlation or mutual information with the outputs. The GP technique as described gives predictive errors equivalent to, if not better than, more widespread methods such as partial least squares and artificial neural networks but additionally can provide a means for easing the interpretation of the correlation between input and output variables. The described application demonstrates that by using the GP method for analyzing PyMS data the adulteration of orange juice with 10% sucrose solution can be quantified reliably over a 0-20% range with an RMS error in the estimate of ∼1%.

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.

Efficient Improvement of Silage Additives by Using Genetic Algorithms

ABSTRACT The enormous variety of substances which may be added to forage in order to manipulate and improve the ensilage process presents an empirical, combinatorial optimization problem of great complexity. To investigate the utility of genetic algorithms for designing effective silage additive combinations, a series of small-scale proof of principle silage experiments were performed with fresh ryegrass. Having established that significant biochemical changes occur over an ensilage period as short as 2 days, we performed a series of experiments in which we used 50 silage additive combinations (prepared by using eight bacterial and other additives, each of which was added at six different levels, including zero [i.e., no additive]). The decrease in pH, the increase in lactate concentration, and the free amino acid concentration were measured after 2 days and used to calculate a “fitness” value that indicated the quality of the silage (compared to a control silage made without additives). This analysis also included a “cost” element to account for different total additive levels. In the initial experiment additive levels were selected randomly, but subsequently a genetic algorithm program was used to suggest new additive combinations based on the fitness values determined in the preceding experiments. The result was very efficient selection for silages in which large decreases in pH and high levels of lactate occurred along with low levels of free amino acids. During the series of five experiments, each of which comprised 50 treatments, there was a steady increase in the amount of lactate that accumulated; the best treatment combination was that used in the last experiment, which produced 4.6 times more lactate than the untreated silage. The additive combinations that were found to yield the highest fitness values in the final (fifth) experiment were assessed to determine a range of biochemical and microbiological quality parameters during full-term silage fermentation. We found that these combinations compared favorably both with uninoculated silage and with a commercial silage additive. The evolutionary computing methods described here are a convenient and efficient approach for designing silage additives.

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.

The detection of caffeine in a variety of beverages using Curie-point pyrolysis mass spectrometry and genetic programming

Freeze dried coffee, filter coffee, tea and cola were analysed by Curie-point pyrolysis mass spectrometry (PyMS). Cluster analysis showed, perhaps not surprisingly, that the discrimination between coffee, tea and cola was very easy. However, cluster analysis also indicated that there was a secondary difference between these beverages which could be attributed to whether they were caffeine-containing or decaffeinated. Artificial neural networks (ANNs) could be trained, with the pyrolysis mass spectra from some of the freeze dried coffees, to classify correctly the caffeine status of the unseen spectra of freeze dried coffee, filter coffee, tea and cola in an independent test set. However, the information in terms of which masses in the mass spectrum are important was not available, which is why ANNs are often perceived as a ‘black box’ approach to modelling spectra. By contrast, genetic programs (GPs) could also be used to classify correctly the caffeine status of the beverages, but which evolved function trees (or mathematical rules) enabling the deconvolution of the spectra and which highlighted that m/z 67, 109 and 165 were the most significant masses for this classification. Moreover, the chemical structure of these mass ions could be assigned to the reproducible pyrolytic degradation products from caffeine.