References

Prediction of obesity based on plasma lipid pattern

In the clinical practice lipid levels, such as cholesterol and triglyceride, are assessed routinely from blood samples, and physicians regularly prescribe lipid-lowering drugs to patients with dysregulated lipid levels. The number of these patients is continuously growing. Dysregulated lipid levels (high trigylceride and/or low HDL cholesterol levels) are metabolic risk factors that rise the risk for heart disease, diabetes, and stroke.
Besides cholesterol and triglycerides, the human plasma contains several hundres to thousand distinct lipid molecular species with outstanding structural diversity. How this striking diversity contribute to the response of a drug therapy is still largely unknown. Therefore, there is a compelling demand for more detailed lipid analyses both for diagnostic purposes and for monitoring the efficacy of prescribed therapy.
 
We developed an efficient, precise and fast mass spectrometric method for the lipidomic screening of human plasma samples. Based on high mass accuracy survey measurements we quantified more than 200 lipid species within 10 minutes. The established analysis is robust, and allows the screen-type examination of high number of samples.
The power of the method was tested on control versus obese sample set.
We could show several significant changes between the lean and overweighted patients. Very importantly, the lipid species that were found significantly altered represented almost all lipid classes analysed, not only triglycerides.
Moreover we found that special pattern of lipid species can characterize much better the pathological states than a few individual lipids could do.
For data analysis a wide range of statistical methods is available By applying the appropriate method, numerous valuable information can be extracted from the raw dataset that make the measurements interpretable, e.g. one can identify marker lipids and categorize patient groups. As an example, the neural network multilayer perceptron model allowed the prediction of obesity based on plasma lipid pattern. Lipidomics with machine learning algorithms can be the future tool for prognosis estimation of either diseases or the therapeutic efficacy of drug candidates.