We then examined whether specific proteins, as opposed to all the proteins in general, were consistently highest or lowest in a certain preparation type. Evidence for such a bias would be indicated by multiple specimen sets showing agreement in the alteration of the concentration of a specific protein, e.g., if all three of the BD specimen sets showed a certain protein higher in a certain preparation method. We identified the proteins that always gave a highest value in one particular preparation method, in every specimen set, and in every replicate experiment. Such biases toward a particular preparation method are more than 99% likely not to have occurred by chance, as determined by a permutation test similar to that described above. A summary of these results is shown in Tab. 1. Many proteins were always highest in serum or in EDTA-plasma, particularly in the DB and GNF sets, respectively. Twenty-four proteins were always lowest in citrate-plasma in the DB set. These specific biases tend to follow the trends seen in the overall concentrations shown in Figs. 3, 4, but occasionally proteins are altered counter to those trends.
For example, in the GNF set, all the four preparation methods had proteins that were consistently elevated. A complete list of the proteins that seem to have concentrations systematically affected by preparation method, along with the magnitudes of the alterations, is provided in the Supplemental Tab. 3. The magnitude of the difference between preparation methods was usually below three-fold, but some proteins had much larger alterations (a ten-fold change or more) in certain preparation methods. The most consistent differences were in the DB set; the 24 proteins that were always lowest in citrate-plasma ranged from 73 to 88% of the maximum values, and the ten proteins that were always highest in serum ranged from 138 to 119% of the minimum values.
The bias for a particular preparation method in a specific protein is visually depicted in Fig. 5 for two representative proteins from each data set. The replicate measurements from each sample were plotted with respect to preparation method, with the solid lines representing the averages between the replicates. In each case shown, one preparation method is consistently highest in every sample and every replicate. EDTA-plasma, heparin-plasma, and serum each have examples in which the concentrations seem to be systematically elevated in one preparation method. Independently-collected ELISA data are plotted along with the microarray data for hemoglobin (Fig. 5G). The concordance between the microarray ELISA measurements are very good for each sample (0.94 over all the samples), validating the accuracy of the microarray measurements and the fact that the hemoglobin concentrations are highest in EDTA-plasma for these samples.
Was this article helpful?