Artificial neural network for monitoring the antioxidant status of human plasma

We evaluated theperformance of a mathematical method to predict the antioxidant power of plasma and the importance of oxidative parameters in humanplasma. One hundred sixty-five blood samples from donors were analyzed in this experimental study. Age, weight, and sex were determined as demographic parameters. Albumin, creatinine, FBS, triglyceride, uric acid and Hb absorbances at 280 to700 nm were analyzed as biochemical parameters. The ferric reducing ability of plasma (FRAP) and carbonyl content of proteins (PCO) were calculated as oxidative markers. An artificial neural network (ANN) was developed as a multilayerfeedforward architecture using IBM SPSS statistics. The best ANN model was performed by a four-layer perceptron method (19-10-10-1) with hyperbolic tangent and identity activation functions for hidden and output layers, respectively. A significant positive correlation (R2 =0.912) was observed between predicted and observed values of FRAP. According to the normalized importance, the main parameters were uric acid (100%), oxyHb (66.8%), A560 (65%), BUN (55%), A420 (52.9%) and creatinine (51.2%). This study demonstrated the ability of the ANN to predict the most important oxidative markers in human plasma. Identification of important parameters can eliminate less important parameters from laboratory procedures and performs a cheaper and faster experiments

کلیدواژه ها:

Antioxidant status ، Artificial neural network ، Human plasma