Many diseases, such as atherosclerosis, type 2 diabetes, the metabolic syndrome and even Alzheimer's disease, which are prevalent in the Western culture, are related to distractions in lipid and lipoprotein metabolism. The urge to gain understanding and even control of these diseases has been a great motivator for metabonomics research - an emerging field, which lays ground for this thesis.

Metabonomics is defined as quantitative observation of metabolic responses to (patho)physiological stimuli or genetic modification in a biological system. Here, the quantitative observations relate to metabolites in body fluids or tissues.

In this work, the focus is on detecting metabolites in human serum by 1H NMR spectroscopy. Getting biological information out of serum spectra can be done by means of mathematical line shape fitting. Traditional way of manually fitting the model for every sample has become too tedious and slow as the sample sizes in metabonomics studies have dramatically increased in recent years.

In this thesis, a software for automated line shape fitting analysis is introduced. The software includes additional functionality for handling collections of spectral data and combining them with the related biological metadata and finally, some tools for quick visualisation of the results. The software is then used to develop a line shape fitting model to quantify metabolites from (particularly) 1H NMR spectra of serum. Finally, the model is tested with a validation dataset.

A great challenge is set on societies and individuals in the near decades as the number of Alzheimer's disease (AD) patients increases dramatically. No medication exists that could reverse, halt or significantly slow down AD progression. Therefore, the focus in AD research has now shifted to patients with mild cognitive impairment (MCI), a condition in which cognitive skills are impaired more than typical to normal aging yet not as severely as in dementia. MCI patients are at high risk of subsequently developing dementia, thus, by studying these individuals it is possible to identify profiles that relate to increased risk of AD.

Recent evidence suggests a major role for cholesterol and lipoproteins in the AD pathology. Also several vascular risk factors associated with altered lipid metabolism are now identified as risk factors for AD. As with 1H NMR metabonomics of serum lipids and lipoproteins can be studied efficiently, this methodology is an appealing approach to study the early stages of AD. Thus far 1H NMR metabonomics has not yet been applied to study cognitive decline and, in addition, other applications on human disease diagnostics and risk assessment are sparse.

We applied 1H NMR metabonomics of serum with a novel combination of three molecular windows and self-organising map analysis to study the metabolic characteristics of MCI. Our results suggest strong connections between MCI and many characteristics associated with the metabolic syndrome, obesity, diabetes, low serum high density lipoprotein cholesterol, high serum triglycerides, coronary artery disease and hypertension. The combination of these unfavourable characteristics related to the highest prevalence of MCI.

The results support the hypothesis that lipid metabolism has a pivotal role in the development of cognitive disorders. Moreover, this study further demonstrates the general applicability of 1H NMR spectroscopy of serum to characterise metabolic phenotypes indicative of increased risk for disease.

Alzheimer's disease (AD) is the most common dementia in the world. It is affecting up to 25 million individuals worldwide and it will double in twenty years. AD is a progressive brain disorder that gradually destroys a person's memory and other cognitive abilities. The initiation and development of AD are poorly understood and there are no distinct biomarkers allowing for early detection and preventive treatment. The major focus in AD research has traditionally been in proteomics but in recent years there has been an increasing interest in the association of serum cholesterol (TC) and lipoproteins with the development of AD.

AD leads to nerve cell death and tissue loss throughout the brain. Plaques and tangles are prime suspects. Plaque form when Amyloid beta (Aβ) peptides clump together. Tangles are formed by phosphorylation of protein named tau, which is promoted by Aβ. Cholesterol is the main component of cell membranes and exists in myelin sheath of nerve cell axons. Cholesterol level affects the formation of the Aβ and thus modulates the production of plaques and tangles. Cholesterol lowering drugs, statins, which are widely used for the treatment of atherosclerosis, are reported to decrease AD progression. Both higher as well as lower TC levels have been indicated to associate with the development of AD. However, only very limited data is available on the associated serum lipoprotein profiles of the AD patients.

The aim of this study is to investigate the role of lipoprotein and cholesterol in AD. AD starts to progress 10 to 20 years before any symptoms can be detected with currently used methods. Therefore early foundation of the risk profiles could help to block the progression of the disease. PubMed searches of English-language articles were used to find all articles in the field. The literature study was made according to these articles. Moreover, information on all studies that included serum and cerebrospinal fluid measurements (CSF) was gathered from the articles found. Results were analysed using statistical meta-analysis in the case of studies that included necessary information.

The results of the comprehensive meta-analysis of the TC studies (between years 1986-2007; n~9000 individuals) reveal that the opposite findings concerning the role of TC and the development of AD can be explained by two distinct lipoprotein profiles: 1) a "conventional" atherosclerotic risk profile (high TC and LDL-C) and 2) a profile associated with the metabolic syndrome (decreased TC and HDL-C plus elevated triglycerides). These results propose that disturbed lipoprotein metabolism is related to the development of AD. This leads to a suggestion that researchers should not focus on discovering only TC levels, but measure more comprehensive data on lipids and lipoproteins.

Atherosclerotic cardiovascular disease is a complex systemic disease in which cholesterol and other lipids are accumulating within the walls of arterial blood vessels. The disease is the cause of heart attacks, stroke, aortic aneurysms, and peripheral vascular disease, which together represent the most frequent causes of death in the industrialized world. In the circulation lipids are carried inside particles called lipoproteins. The lipoprotein particles have a major role in the development of atherosclerosis. They can be divided into five main classes; chylomicrons, VLDL, IDL, LDL and HDL. The LDL particles are the main cholesterol carriers in blood plasma, and the LDL cholesterol is currently considered the most important risk factor for atherosclerosis. In most of the clinical studies the LDL cholesterol is estimated using a mathematical equation, the so-called Friedewald formula. In this formula the LDL cholesterol is estimated from the plasma total cholesterol, triglycerides and HDL cholesterol. However, the Friedewald estimate includes also IDL cholesterol. There is increasing evidence that the IDL cholesterol is an important and independent risk factor for atherosclerosis. Thus, it would probably be beneficial if the LDL and IDL cholesterol could be estimated separately.

The goal of this thesis was to create a model that estimates the LDL and IDL cholesterol separately using the input parameters of the Friedewald formula. For this purpose we have a data set from the University of Oulu that includes 1761 samples. Different models were tested including a linear regression model, multilayer perceptron (MLP) neural network models, and models in which the Bayesian approach was applied to the MLP. An MLP model was selected, and the estimations of this model were used to calculate the relative risks for coronary heart disease (CHD) for the IDL, LDL and IDL+LDL cholesterol from another data set from Helsinki Heart Study including 4081 samples. The IDL cholesterol appeared to be an independent risk factor for CHD. As a conclusion, the MLP model provides an improved Friedewald formula plus most importantly separates estimates for the LDL and IDL cholesterol.

Atherosclerosis is a major cause of cardiovascular disease that is the most common cause of death in the Western World. Lipoproteins and more specifically the lipoprotein subclass distribution, i.e., the so-called lipoprotein profile, plays a key role in the development of atherosclerosis. Some subclasses are known to be more important with respect to the risk for atherosclerosis, but the current clinical risk assessment methodology cannot take this properly into account. 1H NMR spectroscopy enables a fast measurement of the lipoprotein profile from a serum sample, and many computational methods have been developed for this purpose. The use of 1H NMR spectroscopy seems to have significant potential also in clinical use. However, the intrinsic capability and accuracy of this approach to identify and quantify the different lipoprotein subclasses has never been clearly evaluated. In this work this task was approached using various computational methods and an extensive, biochemically sound, set of simulated serum spectra. The overall lipoprotein subclass quantification accuracy was found high when i) the spectral variables were selected according to a high correlation to the lipoprotein subclass concentration, and ii) an early stop-MLP committee, i.e., a regularized nonlinear function approximator, was used. There were clear differences in the methodological accuracy achieved for different lipoprotein subclasses. A novel finding was that the non-lipoprotein components in the 1H NMR spectra of serum do not pose any difficulties for the quantification accuracy, but the largest deviations seem to originate from the inter-individual variation in the composition and structure of the lipoprotein subclass particles. The current results can be taken to support the potential role of 1H NMR spectroscopy in the clinical risk assessment of atherosclerosis. However, population based studies will be needed to detail the effects of individual variations and the maximum quantitative information available from different lipoprotein subclasses in the 1H NMR spectra of serum.

Lipoproteins are small particles which transfer hydrophobic lipids in circulation for cells' energy production or storage. They have variety of sizes, densities, lipid compositions and protein contents by which they are classified as chylomicrons (CM), very-low density lipoproteins (VLDL), intermediate density lipoproteins (IDL), low density lipoproteins (LDL) and high density lipoproteins (HDL). Conventionally, the HDL particles are further divided into HDL2 and HDL3 particles due to their metabolically important differences. Lipoproteins contain several kinds of lipids such as triglyceride, cholesterol ester, phospholipid and free cholesterol molecules. Rest of the lipoprotein particles are proteins called apolipoproteins, which mainly determine the metabolism of the lipoproteins. Although, the structure of the lipoproteins has been studied for roughly half a century and several geometric models have been suggested in the past thirty years, the molecular details and interactions are still not well understood.

At physiological temperatures lipoproteins are mainly spherical particles having diameters from only some nanometres up to several hundreds of nanometres. Their structure can be divided into two regions: a core and about two nanometres thick layer surrounding the core. Since phospholipids, free cholesterol molecules and proteins have both hydrophilic and hydrophobic portions, they locate at the surface of the lipoprotein particles being in touch with the water-based environment on one side and with the interiors of the particles on another side. In addition, small portion of free cholesterol molecules have also been noticed to reside in the core of the lipoprotein particles. It has been generally assumed for decades that all triglycerides and cholesterol esters locate in the core of the lipoprotein particles due to their hydrophobic nature. However, it is metabolically plausible that a portion of these lipids reside in the surface layer of the lipoprotein particles.

This thesis presents a new approach to determine lipid distributions in lipoprotein particles. In the model each lipoprotein particle is modelled as a sphere which is divided into a hydrophobic core and a surrounding surface layer resembling real particles. It is also assumed that all phospholipids and proteins reside at the surface according to the current data and understanding. The purpose was to determine the relative portion of triglyceride, cholesterol ester and free cholesterol molecules in the core (and consequently in the surface layer) using constrained optimisation. The optimisation procedure used a combination of a simple Monte Carlo technique and standard minimisation algorithm to find the global minimum of an error function. It was shown using simulated data that the method was able to find the distributions even from noisy data. However, it was noted that a large amount of data was required to compensate the noise.

Having shown that the optimisation procedure was sufficient, the model was applied to experimental data that consisted of measured lipid and protein concentrations for VLDL, IDL, LDL, HDL2 and HDL3 particles from approximately 300 individuals. The optimisation procedure found well defined distributions to triglycerides and cholesterol esters but the distribution of free cholesterol molecules were undetermined. The results clearly indicate that remarkable portions of triglycerides and cholesterol esters locate in the surface layer of lipoprotein particles: the smaller the lipoprotein particles the higher were also the distribution of hydrophobic "core" lipids in the surface.