|
Abstract:
|
Mass spectrometry is used widely to determine protein functions in living organs. Extracting useful information is a key point. There is no standard for measuring which extraction methodology gives the best result. We proposed a methodology to preprocess the spectrum data using misclassification rate as a rule to evaluate and compare preprocessing methods. In this paper, an objective function of baseline correction is proposed, subject to two constraints. After necessary adjustments of the data, our approach uses a nested nearest-neighbor classification scheme, applying three commonly used distance measures: correlation between patients, Euclidean metric, and Canberra metric. Our finding shows our preprocessing method yields a plausible rate of misclassification rates and a better result than previous studies.
|
