A. VĂLEANU#, MIHAELA ILIE, INES DIMA, CARMEN PURDEL
Toxicology Department, Faculty of Pharmacy, “Carol Davila” University of Medicine and Pharmacy, Bucharest, Romania
Protein carbonyls are very important biomarkers of oxidative stress. Pattern recognition (PR) methods use certain mathematical algorithms to derive qualitative features of samples based on their parameters. The aim of the paper was to find a pattern in the electrophoregrams of carbonyl proteins from human serum albumin using PR techniques. Samples of previously oxidized human serum albumin (HA) were analyzed using capillary electrophoresis (CE) at 214 nm and 365 nm. Five samples were selected to build the pattern and several peak classes were created at the two selected wavelengths using cluster analysis. Several serum samples from diabetes patients were also analyzed and compared to those of the master samples using the k nearest neighbours method (kNN). The preliminary results indicate that a series of CE peaks can be found in most of the HA sample runs, with small variability in each peak class. Only a few of the patients’ peaks were found to fit the pattern, with retention time being the most similar parameter. KNN and cluster analysis are both simple, yet efficient techniques, capable of a complex and profound analysis of the studied system and can be used to build the pattern of carbonylated human serum albumin.
Corresponding author’s e-mail: firstname.lastname@example.orgFull text: PDF