Characterization of 1H NMR spectroscopic data and the generation of synthetic validation sets.

MOTIVATION: Common contemporary practice within the nuclear magnetic resonance (NMR) metabolomics community is to evaluate and validate novel algorithms on empirical data or simplified simulated data. Empirical data captures the complex characteristics of experimental data, but the optimal or most correct analysis is unknown a priori; therefore, researchers are forced to rely on indirect performance metrics, which are of limited value. In order to achieve fair and complete analysis of competing techniques more exacting metrics are required. Thus, metabolomics researchers often evaluate their algorithms on simplified simulated data with a known answer. Unfortunately, the conclusions obtained on simulated data are only of value if the data sets are complex enough for results to generalize to true experimental data. Ideally, synthetic data should be indistinguishable from empirical data, yet retain a known best analysis. RESULTS: We have developed a technique for creating realistic synthetic metabolomics validation sets based on NMR spectroscopic data. The validation sets are developed by characterizing the salient distributions in sets of empirical spectroscopic data. Using this technique, several validation sets are constructed with a variety of characteristics present in 'real' data. A case study is then presented to compare the relative accuracy of several alignment algorithms using the increased precision afforded by these synthetic data sets. AVAILABILITY: These data sets are available for download at http://birg.cs.wright.edu/nmr_synthetic_data_sets.

Knowledge discovery in medical and biological datasets using a hybrid Bayes classifier/evolutionary algorithm.

A key element of bioinformatics research is the extraction of meaningful information from large experimental data sets. Various approaches, including statistical and graph theoretical methods, data mining, and computational pattern recognition, have been applied to this task with varying degrees of success. Using a novel classifier based on the Bayes discriminant function, we present a hybrid algorithm that employs feature selection and extraction to isolate salient features from large medical and other biological data sets. We have previously shown that a genetic algorithm coupled with a k-nearest-neighbors classifier performs well in extracting information about protein-water binding from X-ray crystallographic protein structure data. The effectiveness of the hybrid EC-Bayes classifier is demonstrated to distinguish the features of this data set that are the most statistically relevant and to weight these features appropriately to aid in the prediction of solvation sites.