Teemu Suna

Quantitative Serum Nuclear Magnetic Resonance Metabolomics in Cardiovascular Epidemiology and Genetics

Metabolomics is becoming common in epidemiology due to recent developments in quantitative profiling technologies and appealing results from their applications for understanding health and disease. Our team has developed an automated high-throughput serum NMR metabolomics platform that provides quantitative molecular data on 14 lipoprotein subclasses, their lipid concentrations and composition, apolipoprotein A-I and B, multiple cholesterol and triglyceride measures, albumin, various fatty acids as well as on numerous low-molecular-weight metabolites, including amino acids, glycolysis related measures and ketone bodies. The molar concentrations of these measures are obtained from a single serum sample with costs comparable to standard lipid measurements. We have analyzed almost 250 000 samples from around 100 epidemiological cohorts and biobanks and the new international set-up of multiple platforms will allow an annual throughput of more than 250 000 samples. The molecular data have been used to study type 1 and type 2 diabetes etiology as well as to characterize the molecular reflections of the metabolic syndrome, long-term physical activity, diet and lipoprotein metabolism. The results have revealed new biomarkers for early atherosclerosis, type 2 diabetes, diabetic nephropathy, cardiovascular disease and all-cause mortality. We have also combined genomics and metabolomics in diverse studies. We envision that quantitative high-throughput NMR metabolomics will be incorporated as a routine in large biobanks; this would make perfect sense both from the biological research and cost point of view - the standard output of over 200 molecular measures would vastly extend the relevance of the sample collections and make many separate clinical chemistry assays redundant.

Self-Organising Map Approach to Individual Profiles: Age, Sex and Culture in Internet Dating

A marked feature of recent developments in the networked society has been the growth in the number of people making use of Internet dating services. These services involve the accumulation of large amounts of personal information which individuals utilise to find others and potentially arrange offline meetings. The consequent data represent a challenge to conventional analysis, for example, the service that provided the data used in this paper had approximately 5,000 users all of whom completed an extensive questionnaire resulting in some 300 parameters. This creates an opportunity to apply innovative analytical techniques that may provide new sociological insights into complex data. In this paper we utilise the self-organising map (SOM), an unsupervised neural network methodology, to explore Internet dating data. The resulting visual maps are used to demonstrate the ability of SOMs to reveal interrelated parameters. The SOM process led to the emergence of correlations that were obscured in the original data and pointed to the role of what we call ‘cultural age’ in the profiles and partnership preferences of the individuals. Our results suggest that the SOM approach offers a well established methodology that can be easily applied to complex sociological data sets. The SOM outcomes are discussed in relation to other research about identifying others and forming relationships in a network society.