Data analysis is only as successful as the quality of the input data. A study of approximately 400 thoroughbred horses was undertaken to look for markers of the disease osteochondrosis dessicans (OCD), with the additional aim of determining a genetic basis for the condition.
There is a > 30% risk of OCD in certain breeds of horses, with a heritable predisposition identified in three breeds. The presence of lesions results in considerable economic losses, from surgical correction and also a reduction in value and performance.1 The actual genetic defect causing OCD has not been identified, nor have the metabolic pathways affected in the development of lesions, although previous work has linked hyperinsulinemia in OCD-affected horses.2 Preliminary metabonomic studies1 revealed differences in various low-density lipoproteins and very low density lipoproteins along with choline, which contributed significantly to the clustering of young Standardbreds that developed OCD versus those that did not.
Fasting was not controlled at the time of sampling, nor was exercise prior to taking the serum sample. Finally, inconsistent sampling techniques resulted in a significant number of samples being contaminated by red blood cells. All three of these factors have contributed to a data set which presented difficulties prior to obtaining the NMR spectra.
A second major area where problems can arise is sample preparation. Blood serum samples will generally contain some level of lipids and/or triglycerides. If not treated in some fashion, these fatty components can significantly hinder the NMR acquisition. Three methods exist to remove these components. First is the use of sub-micron filtration, next organic extraction of the aqueous solution to remove the fats, and finally spectroscopic elimination using a CPMG pulse sequence or similar.