Automatic learning for the classification of chemical reactions and in statistical thermodynamics

This Thesis describes the application of automatic learning methods for a) the classification of organic and metabolic reactions, and b) the mapping of Potential Energy Surfaces(PES). The classification of reactions was approached with two distinct methodologies: a representation of chemical reactions based on NMR data, and a representation of chemical reactions from the reaction equation based on the physico-chemical and topological features of chemical bonds. NMR-based classification of photochemical and enzymatic reactions. Photochemical and metabolic reactions were classified by Kohonen Self-Organizing Maps (Kohonen SOMs) and Random Forests (RFs) taking as input the difference between the 1H NMR spectra of the products and the reactants. The development of such a representation can be applied in automatic analysis of changes in the 1H NMR spectrum of a mixture and their interpretation in terms of the chemical reactions taking place. Examples of possible applications are the monitoring of reaction processes, evaluation of the stability of chemicals, or even the interpretation of metabonomic data. A Kohonen SOM trained with a data set of metabolic reactions catalysed by transferases was able to correctly classify 75% of an independent test set in terms of the EC number subclass. Random Forests improved the correct predictions to 79%. With photochemical reactions classified into 7 groups, an independent test set was classified with 86-93% accuracy. The data set of photochemical reactions was also used to simulate mixtures with two reactions occurring simultaneously. Kohonen SOMs and Feed-Forward Neural Networks (FFNNs) were trained to classify the reactions occurring in a mixture based on the 1H NMR spectra of the products and reactants. Kohonen SOMs allowed the correct assignment of 53-63% of the mixtures (in a test set). Counter-Propagation Neural Networks (CPNNs) gave origin to similar results. The use of supervised learning techniques allowed an improvement in the results. They were improved to 77% of correct assignments when an ensemble of ten FFNNs were used and to 80% when Random Forests were used. This study was performed with NMR data simulated from the molecular structure by the SPINUS program. In the design of one test set, simulated data was combined with experimental data. The results support the proposal of linking databases of chemical reactions to experimental or simulated NMR data for automatic classification of reactions and mixtures of reactions. Genome-scale classification of enzymatic reactions from their reaction equation. The MOLMAP descriptor relies on a Kohonen SOM that defines types of bonds on the basis of their physico-chemical and topological properties. The MOLMAP descriptor of a molecule represents the types of bonds available in that molecule. The MOLMAP descriptor of a reaction is defined as the difference between the MOLMAPs of the products and the reactants, and numerically encodes the pattern of bonds that are broken, changed, and made during a chemical reaction. The automatic perception of chemical similarities between metabolic reactions is required for a variety of applications ranging from the computer validation of classification systems, genome-scale reconstruction (or comparison) of metabolic pathways, to the classification of enzymatic mechanisms. Catalytic functions of proteins are generally described by the EC numbers that are simultaneously employed as identifiers of reactions, enzymes, and enzyme genes, thus linking metabolic and genomic information. Different methods should be available to automatically compare metabolic reactions and for the automatic assignment of EC numbers to reactions still not officially classified. In this study, the genome-scale data set of enzymatic reactions available in the KEGG database was encoded by the MOLMAP descriptors, and was submitted to Kohonen SOMs to compare the resulting map with the official EC number classification, to explore the possibility of predicting EC numbers from the reaction equation, and to assess the internal consistency of the EC classification at the class level. A general agreement with the EC classification was observed, i.e. a relationship between the similarity of MOLMAPs and the similarity of EC numbers. At the same time, MOLMAPs were able to discriminate between EC sub-subclasses. EC numbers could be assigned at the class, subclass, and sub-subclass levels with accuracies up to 92%, 80%, and 70% for independent test sets. The correspondence between chemical similarity of metabolic reactions and their MOLMAP descriptors was applied to the identification of a number of reactions mapped into the same neuron but belonging to different EC classes, which demonstrated the ability of the MOLMAP/SOM approach to verify the internal consistency of classifications in databases of metabolic reactions. RFs were also used to assign the four levels of the EC hierarchy from the reaction equation. EC numbers were correctly assigned in 95%, 90%, 85% and 86% of the cases (for independent test sets) at the class, subclass, sub-subclass and full EC number level,respectively. Experiments for the classification of reactions from the main reactants and products were performed with RFs - EC numbers were assigned at the class, subclass and sub-subclass level with accuracies of 78%, 74% and 63%, respectively. In the course of the experiments with metabolic reactions we suggested that the MOLMAP / SOM concept could be extended to the representation of other levels of metabolic information such as metabolic pathways. Following the MOLMAP idea, the pattern of neurons activated by the reactions of a metabolic pathway is a representation of the reactions involved in that pathway - a descriptor of the metabolic pathway. This reasoning enabled the comparison of different pathways, the automatic classification of pathways, and a classification of organisms based on their biochemical machinery. The three levels of classification (from bonds to metabolic pathways) allowed to map and perceive chemical similarities between metabolic pathways even for pathways of different types of metabolism and pathways that do not share similarities in terms of EC numbers. Mapping of PES by neural networks (NNs). In a first series of experiments, ensembles of Feed-Forward NNs (EnsFFNNs) and Associative Neural Networks (ASNNs) were trained to reproduce PES represented by the Lennard-Jones (LJ) analytical potential function. The accuracy of the method was assessed by comparing the results of molecular dynamics simulations (thermal, structural, and dynamic properties) obtained from the NNs-PES and from the LJ function. The results indicated that for LJ-type potentials, NNs can be trained to generate accurate PES to be used in molecular simulations. EnsFFNNs and ASNNs gave better results than single FFNNs. A remarkable ability of the NNs models to interpolate between distant curves and accurately reproduce potentials to be used in molecular simulations is shown. The purpose of the first study was to systematically analyse the accuracy of different NNs. Our main motivation, however, is reflected in the next study: the mapping of multidimensional PES by NNs to simulate, by Molecular Dynamics or Monte Carlo, the adsorption and self-assembly of solvated organic molecules on noble-metal electrodes. Indeed, for such complex and heterogeneous systems the development of suitable analytical functions that fit quantum mechanical interaction energies is a non-trivial or even impossible task. The data consisted of energy values, from Density Functional Theory (DFT) calculations, at different distances, for several molecular orientations and three electrode adsorption sites. The results indicate that NNs require a data set large enough to cover well the diversity of possible interaction sites, distances, and orientations. NNs trained with such data sets can perform equally well or even better than analytical functions. Therefore, they can be used in molecular simulations, particularly for the ethanol/Au (111) interface which is the case studied in the present Thesis. Once properly trained, the networks are able to produce, as output, any required number of energy points for accurate interpolations.

Production of chitin-glucan complex by Pichia pastoris

Dissertation for the Degree of Master in Biotechnology

Experimental and modeling studies on reverse electrodialysis for sustainable power generation

A potentially renewable and sustainable source of energy is the chemical energy associated with solvation of salts. Mixing of two aqueous streams with different saline concentrations is spontaneous and releases energy. The global theoretically obtainable power from salinity gradient energy due to World’s rivers discharge into the oceans has been estimated to be within the range of 1.4-2.6 TW. Reverse electrodialysis (RED) is one of the emerging, membrane-based, technologies for harvesting the salinity gradient energy. A common RED stack is composed by alternately-arranged cation- and anion-exchange membranes, stacked between two electrodes. The compartments between the membranes are alternately fed with concentrated (e.g., sea water) and dilute (e.g., river water) saline solutions. Migration of the respective counter-ions through the membranes leads to ionic current between the electrodes, where an appropriate redox pair converts the chemical salinity gradient energy into electrical energy. Given the importance of the need for new sources of energy for power generation, the present study aims at better understanding and solving current challenges, associated with the RED stack design, fluid dynamics, ionic mass transfer and long-term RED stack performance with natural saline solutions as feedwaters. Chronopotentiometry was used to determinate diffusion boundary layer (DBL) thickness from diffusion relaxation data and the flow entrance effects on mass transfer were found to avail a power generation increase in RED stacks. Increasing the linear flow velocity also leads to a decrease of DBL thickness but on the cost of a higher pressure drop. Pressure drop inside RED stacks was successfully simulated by the developed mathematical model, in which contribution of several pressure drops, that until now have not been considered, was included. The effect of each pressure drop on the RED stack performance was identified and rationalized and guidelines for planning and/or optimization of RED stacks were derived. The design of new profiled membranes, with a chevron corrugation structure, was proposed using computational fluid dynamics (CFD) modeling. The performance of the suggested corrugation geometry was compared with the already existing ones, as well as with the use of conductive and non-conductive spacers. According to the estimations, use of chevron structures grants the highest net power density values, at the best compromise between the mass transfer coefficient and the pressure drop values. Finally, long-term experiments with natural waters were performed, during which fouling was experienced. For the first time, 2D fluorescence spectroscopy was used to monitor RED stack performance, with a dedicated focus on following fouling on ion-exchange membrane surfaces. To extract relevant information from fluorescence spectra, parallel factor analysis (PARAFAC) was performed. Moreover, the information obtained was then used to predict net power density, stack electric resistance and pressure drop by multivariate statistical models based on projection to latent structures (PLS) modeling. The use in such models of 2D fluorescence data, containing hidden, but extractable by PARAFAC, information about fouling on membrane surfaces, considerably improved the models fitting to the experimental data.

The thermal effects on the methanol-to-olefins reaction: A modelling and experimental approach

With the projection of an increasing world population, hand-in-hand with a journey towards a bigger number of developed countries, further demand on basic chemical building blocks, as ethylene and propylene, has to be properly addressed in the next decades. The methanol-to-olefins (MTO) is an interesting reaction to produce those alkenes using coal, gas or alternative sources, like biomass, through syngas as a source for the production of methanol. This technology has been widely applied since 1985 and most of the processes are making use of zeolites as catalysts, particularly ZSM-5. Although its selectivity is not especially biased over light olefins, it resists to a quick deactivation by coke deposition, making it quite attractive when it comes to industrial environments; nevertheless, this is a highly exothermic reaction, which is hard to control and to anticipate problems, such as temperature runaways or hot-spots, inside the catalytic bed. The main focus of this project is to study those temperature effects, by addressing both experimental, where the catalytic performance and the temperature profiles are studied, and modelling fronts, which consists in a five step strategy to predict the weight fractions and activity. The mind-set of catalytic testing is present in all the developed assays. It was verified that the selectivity towards light olefins increases with temperature, although this also leads to a much faster catalyst deactivation. To oppose this effect, experiments were carried using a diluted bed, having been able to increase the catalyst lifetime between 32% and 47%. Additionally, experiments with three thermocouples placed inside the catalytic bed were performed, analysing the deactivation wave and the peaks of temperature throughout the bed. Regeneration was done between consecutive runs and it was concluded that this action can be a powerful means to increase the catalyst lifetime, maintaining a constant selectivity towards light olefins, by losing acid strength in a steam stabilised zeolitic structure. On the other hand, developments on the other approach lead to the construction of a raw basic model, able to predict weight fractions, that should be tuned to be a tool for deactivation and temperature profiles prediction.

Cell functional enviromics: applications to Pichia pastoris

The present PhD thesis develops the cell functional enviromics (CFE) method to investigate the relationship between environment and cellular physiology. CFE may be defined as the envirome-wide cellular function reconstruction through the collection and systems-level analysis of dynamic envirome data. Throughout the thesis, CFE is illustrated by two main applications to cultures of a constitutive P. pastoris X33 strain expressing a scFv antibody fragment. The first application addresses the challenge of culture media development. A dataset was built from 26 shake flask experiments, with variations in trace elements concentrations and basal medium dilution based on the standard BSM+PTM1. Protein yield showed high sensitivity to culture medium variations, while biomass was essentially determined by BSM dilution. High scFv yield was associated with high overall metabolic fluxes through central carbon pathways concomitantly with a relative shift of carbon flux from biosynthetic towards energy-generating pathways. CFE identified three cellular functions (growth, energy generation and by-product formation) that together described 98.8% of the variance in observed fluxes. Analyses of how medium factors relate to identified cellular functions showed iron and manganese at concentrations close to PTM1 inhibit overall metabolic activity. The second application addresses bioreactor operation. Pilot 50 L fed-batch cultivations, followed by 1H-NMR exometabolite profiling, allowed the acquisition of data for 21 environmental factors over time. CFE identified five major metabolic pathway groups that are frequently activated by the environment. The resulting functional enviromics map may serve as template for future optimization of media composition and feeding strategies for Pichia pastoris. The present PhD thesis is a step forward towards establishing the foundations of CFE that is still at its infancy. The methods developed herein are a contribution for changing the culture media and process development paradigm towards a holistic and systematic discipline in the future.

Methodology for WSN communication technologies automated field tests

Wireless Sensor Networks(WSN) are networks of devices used to sense and act that applies wireless radios to communicate. To achieve a successful implementation of a wireless device it is necessary to take in consideration the existence of a wide variety of radios available, a large number of communication parameters (payload, duty cycle, etc.) and environmental conditions that may affect the device’s behaviour. However, to evaluate a specific radio towards a unique application it might be necessary to conduct trial experiments, with such a vast amount of devices, communication parameters and environmental conditions to take into consideration the number of trial cases generated can be surprisingly high. Thus, making trial experiments to achieve manual validation of wireless communication technologies becomes unsuitable due to the existence of a high number of trial cases on the field. To overcome this technological issue an automated test methodology was introduced, presenting the possibility to acquire data regarding the device’s behaviour when testing several technologies and parameters that care for a specific analysis. Therefore, this method advances the validation and analysis process of the wireless radios and allows the validation to be done without the need of specific and in depth knowledge about wireless devices.