Since the beginning of proteomics, the quality of the data has been an important question. Furthermore, the predictive value of omics data is also an important point to be taken into account for advancing biology. To investigate these questions, a model study on macrophages submitted to treatment by copper ion complexed or not with polyacrylate has been used. In this study, the same protein extracts were used in two different proteomic setups, i.e. 2D-gel based and label-free shotgun proteomics.
The quality of the data was investigated by using several metrics, including clustering techniques. While the 2D gel data produced a clustering tree reflecting the expected biological situation, the shotgun data did not produce a consistent grouping, showing that the noise of the data was not negligible compared to the biological signal.
To address the even more important question of the positive predictive value, a "gold standard" is required. The fact that several enzymes appeared as modulated in the various proteomic setups allowed testing the predictive value against the enzyme activities, used as the gold standard. In this dataset, at a cutoff value of p<0.05, the predictive value of the 2D gel data was slightly higher than 40%, while the one of the shotgun data was <20%. It increased at 25% with a cutoff value of p<0.01, and further filtering with the fold change did not induce any improvement, but only a loss in the true positives detected. I the same trend, the calculated FDRs for individual enzymes were not different between the true and false positives.
Although these figures are expected to be variable from one dataset to another, they show that the predictive value of the proteomic data is a figure to be investigated, and that simple enzyme assays are an interesting tool to investigate this parameter.