Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
5371490 | Biophysical Chemistry | 2010 | 7 Pages |
Melting curves of human plasma measured by differential scanning calorimetry (DSC), known as thermograms, have the potential to markedly impact diagnosis of human diseases. A general statistical methodology is developed to analyze and classify DSC thermograms to analyze and classify thermograms. Analysis of an acquired thermogram involves comparison with a database of empirical reference thermograms from clinically characterized diseases. Two parameters, a distance metric, P, and correlation coefficient, r, are combined to produce a 'similarity metric,' Ï, which can be used to classify unknown thermograms into pre-characterized categories. Simulated thermograms known to lie within or fall outside of the 90% quantile range around a median reference are also analyzed. Results verify the utility of the methods and establish the apparent dynamic range of the metric Ï. Methods are then applied to data obtained from a collection of plasma samples from patients clinically diagnosed with SLE (lupus). High correspondence is found between curve shapes and values of the metric Ï. In a final application, an elementary classification rule is implemented to successfully analyze and classify unlabeled thermograms. These methods constitute a set of powerful yet easy to implement tools for quantitative classification, analysis and interpretation of DSC plasma melting curves.
Graphical AbstractDownload full-size imageResearch HighlightsâºResearch Highlights âºStatistical methods for plasma thermogram analysis. âºCharacterization of disease states. âºAnalysis of simulated plasma thermograms. âºClassification of Healthy and SLE (lupus) thermograms.