Proteomic pattern analysis using a neural network application.
Chidambaram-S; Rao-KMK; Ahluwalia-RS
FASEB J 2004 Mar; 18(5)(Suppl):A1220
To determine if exposure to diesel exhaust fumes causes a change in the serum proteomic patterns, serum samples were collected from 80 subjects (46 controls, 34 exposed). Serum proteomic profiles were performed on the Ciphergen Protein Chip® System using WCX2 chips. Proteomic patterns were analyzed by neural network techniques (classification and clustering algorithms) using "Predict" software obtained from Neuralware Inc. The Backpropagation algorithm was used as the classification algorithm and Self-Organizing Maps (SOM) was used as the clustering algorithm. Two methods were used for the identification of the most discriminating peaks. The first method used manual analysis of raw data using Euclidean distance as the criterion and the second method used a p-value statistic obtained from the Ciphergen software. The classification and clustering algorithms were applied to the two data sets. These procedures yielded a sensitivity of 82.5% and specificity of 81% using the peaks selected by the manual data analysis and a sensitivity of 90% and specificity of 92% using the peaks selected by the p-value analysis. These data indicate that a given serum profile pattern can be assigned to diesel exposure group at about 90% confidence limits using a neural network application.
Serological-techniques; Mathematical-models; Analytical-processes; Analytical-chemistry; Analytical-methods; Analytical-models
NIOSH, PPRB, HELD, Morgantown, WV 26505
Abstract; Conference/Symposia Proceedings
The FASEB Journal