An Exploration of Some Factors Affecting the Correlation of mRNA and Proteomic Data

Abstract

The recent availability of technologies for high throughput proteome analysis has led to the emergence of integrated mRNA and protein expression data. In one such study by Ideker and co-workers, changes in mRNA and protein abundance levels were quantified following systematic perturbation of a specific metabolic pathway [1]. The authors calculated an overall Pearson correlation coefficient between changes in mRNA and protein expression of 0.61, however, no change in protein expression was observed for almost 80% of genes reported as having a significant change in mRNA indicating that a complex relationship exists between mRNA and protein expression. To try and address this issue, the data were sorted according to various criteria: protein and mRNA expression ratios, confidence values, length of protein, fraction of cysteine residues and half-life prediction, to try and identify any bias in experimental technique which may affect the correlation. mRNA expression ratio and the confidence value had the strongest affect on how well the data correlated, whilst protein detection was weakly dependent on the fraction of cysteine residues in the protein. Initial investigations have indicated that integrating the data with domain knowledge provides the best opportunity for distinguishing between those transcriptome results which may be interpreted in a straightforward manner and those which should be treated with caution.