Jeremy John Rowland

The Automation of Science

The basis of science is the hypothetico-deductive method and the recording of experiments in sufficient detail to enable reproducibility. We report the development of Robot Scientist “Adam,” which advances the automation of both. Adam has autonomously generated functional genomics hypotheses about the yeast Saccharomyces cerevisiae and experimentally tested these hypotheses by using laboratory automation. We have confirmed Adams conclusions through manual experiments. To describe Adams research, we have developed an ontology and logical language. The resulting formalization involves over 10,000 different research units in a nested treelike structure, 10 levels deep, that relates the 6.6 million biomass measurements to their logical description. This formalization describes how a machine contributed to scientific knowledge.

Rapid and Quantitative Detection of the Microbial Spoilage of Meat by Fourier Transform Infrared Spectroscopy and Machine Learning

ABSTRACT Fourier transform infrared (FT-IR) spectroscopy is a rapid, noninvasive technique with considerable potential for application in the food and related industries. We show here that this technique can be used directly on the surface of food to produce biochemically interpretable “fingerprints.” Spoilage in meat is the result of decomposition and the formation of metabolites caused by the growth and enzymatic activity of microorganisms. FT-IR was exploited to measure biochemical changes within the meat substrate, enhancing and accelerating the detection of microbial spoilage. Chicken breasts were purchased from a national retailer, comminuted for 10 s, and left to spoil at room temperature for 24 h. Every hour, FT-IR measurements were taken directly from the meat surface using attenuated total reflectance, and the total viable counts were obtained by classical plating methods. Quantitative interpretation of FT-IR spectra was possible using partial least-squares regression and allowed accurate estimates of bacterial loads to be calculated directly from the meat surface in 60 s. Genetic programming was used to derive rules showing that at levels of 107 bacteria·g−1 the main biochemical indicator of spoilage was the onset of proteolysis. Thus, using FT-IR we were able to acquire a metabolic snapshot and quantify, noninvasively, the microbial loads of food samples accurately and rapidly in 60 s, directly from the sample surface. We believe this approach will aid in the Hazard Analysis Critical Control Point process for the assessment of the microbiological safety of food at the production, processing, manufacturing, packaging, and storage levels.

Classification of pyrolysis mass spectra by fuzzy multivariate rule induction-comparison with regression, K-nearest neighbour, neural and decision-tree methods

The fuzzy multivariate rule building export system (FuRES) is applied to solve classification problems using two pyrolysis mass spectral data sets. The first data set contains three types of milk (from cow, goat and ewe) and the second data set contains two types of olive oils (adulterated and extra virgin). The performance of FuRES is compared with a selection of well-known classification algorithms: backpropagation artificial neural networks (ANNs), canonical variates analysis (CVA), classification and regression trees (CART), the K-nearest neighbour method (KNN) and discriminant partial least squares (DPLS). In terms of percent correct classification the DPLS and ANNs were best since all test set objects in both data sets were correctly classified. FuRES was second best with 100% correct classification for the milk data set and 91% correct classification for the olive oil data set, while the KNN method showed 100% and 61% for the two data sets. CVA had a 100% correct classification for the milk data set, but failed to form a model for the olive oil data set. The percent correct classifications for the CART method were 92% and 74%, respectively. When both model interpretation and predictive ability are taken into consideration, we consider that the ranking of these methods on the basis of these two data sets is in order of decreasing utility: DPLS, FuRJZS, ANNs, CART, CVA and KNN. Keywonis: Rule induction; Canonical variate analysis; Discriminant partial least squares; PLSZ; K-nearest neighbour method, Fuzzy rule building expert system (FE Artificial neural networks; Classification and regression trees (CART); Pyrolysis mass spectrometry (PyMS)

Discrimination of Modes of Action of Antifungal Substances by Use of Metabolic Footprinting

ABSTRACT Diploid cells of Saccharomyces cerevisiae were grown under controlled conditions with a Bioscreen instrument, which permitted the essentially continuous registration of their growth via optical density measurements. Some cultures were exposed to concentrations of a number of antifungal substances with different targets or modes of action (sterol biosynthesis, respiratory chain, amino acid synthesis, and the uncoupler). Culture supernatants were taken and analyzed for their “metabolic footprints” by using direct-injection mass spectrometry. Discriminant function analysis and hierarchical cluster analysis allowed these antifungal compounds to be distinguished and classified according to their modes of action. Genetic programming, a rule-evolving machine learning strategy, allowed respiratory inhibitors to be discriminated from others by using just two masses. Metabolic footprinting thus represents a rapid, convenient, and information-rich method for classifying the modes of action of antifungal substances.

Model selection methodology in supervised learning with evolutionary computation.

The expressive power, powerful search capability, and the explicit nature of the resulting models make evolutionary methods very attractive for supervised learning applications in bioinformatics. However, their characteristics also make them highly susceptible to overtraining or to discovering chance relationships in the data. Identification of appropriate criteria for terminating evolution and for selecting an appropriately validated model is vital. Some approaches that are commonly applied to other modelling methods are not necessarily applicable in a straightforward manner to evolutionary methods. An approach to model selection is presented that is not unduly computationally intensive. To illustrate the issues and the technique two bioinformatic datasets are used, one relating to metabolite determination and the other to disease prediction from gene expression data.

Generalisation and model selection in supervised learning with evolutionary computation

EC-based supervised learning has been demonstrated to be an effective approach to forming predictive models in genomics, spectral interpretation, and other problems in modern biology. Longer-established methods such as PLS and ANN are also often successful. In supervised learning, overtraining is always a potential problem. The literature reports numerous methods of validating predictive models in order to avoid overtraining. Some of these approaches can be applied to EC-based methods of supervised learning, though the characteristics of EC learning are different from those obtained with PLS and ANN and selecting a suitably general model can be more difficult. This paper reviews the issues and various approaches, illustrating salient points with examples taken from applications in bioinformatics.

The Robot Scientist Adam

Despite sciences great intellectual prestige, developing robot scientists will probably be simpler than developing general AI systems because there is no essential need to take into account the social milieu.

A MODULAR APPROACH TO SENSOR INTEGRATION IN ROBOTIC ASSEMBLY

Abstract We describe the use of functional abstraction for the design of modular sensor integration systems that provide task supervisors with the sensor information required to perform robotic assembly tasks or subtasks. We offer a versatile approach to high-performance sensor integration. This is consistent with the low-cost robotic assembly cells that will be needed to make automated assembly attractive to a wider sector of manufacturing industry.

Nondestructive analysis of senescence in mesophyll cells by spectral resolution of protein synthesis-dependent pigment metabolism

* Over 6 d of dark-induced senescence, leaf segments of wild-type Lolium temulentum lost > 96% chlorophyll a + b; leaves from plants containing a staygreen mutation introgressed from Festuca pratensis, which has a lesion in the senescence-associated fragmentation of pigment-proteolipid complexes, retained over 43% of total chlorophyll over the same period. * Mutant segments preferentially retained thylakoid membrane proteins (exemplified by LHCP II) but lost other cellular proteins at the same rate as wild-type tissue. The protein synthesis inhibitor D-MDMP inhibited chlorophyll degradation and partially prevented protein loss in both genotypes, but tissues treated with the ineffective L-stereoisomer were indistinguishable from water controls. * Principal-components analysis of leaf reflectance spectra distinguished between genotypes, time points and D-MDMP treatments, showing the disruption of pigment metabolism during senescence brought about by the staygreen mutation, by inhibition of protein synthesis and by combinations of the two factors. * The build-up of oxidized, dephytylated and phaeo-derivatives of chl a during senescence of staygreen tissue was prevented by D-MDMP and associated with characteristic difference spectra when senescent mutant tissue was compared with wild-type or inhibitor-treated samples. The suitability of senescence as a subject for systems biology approaches is discussed.

Histometrics: improvement of the dynamic range of fluorescently stained proteins resolved in electrophoretic gels using hyperspectral imaging

Most image‐based analyses, using absorbance or fluorescence of the spatial distribution of identifiable structures in complex biological systems, use only a very small number of dimensions of possible spectral data for the generation and interpretation of the image. We here extend the concepts of hyperspectral imaging, being developed in remote sensing, into analytical biotechnology. The massive volume of information contained in hyperspectral spectroscopic images requires multivariate analysis in order to extract the chemical and spatial information contained within the data. We here describe the use of multivariate statistical methods to map and quantify common protein staining fluorophores (SYPRO Red, Orange and Tangerine) in electrophoretic gels. Specifically, we find (a) that the `background’ underpinning limits of detection is due more to proteins that have not migrated properly than to impurities or to ineffective destaining, (b) the detailed mechanisms of staining of SYPRO red and orange are apparently not identical, and in particular (c) that these methods can provide two orders of magnitude improvement in the detection limit per pixel, to levels well below the limit observable optically.