In Silico Instrumental Response Correction Improves Precision of Label-free Proteomics and Accuracy of Proteomics-based Predictive Models

In the analysis of proteome changes arising during the early stages of a biological process (e.g. disease or drug treatment) or from the indirect influence of an important factor, the biological variations of interest are often small (∼10%). The corresponding requirements for the precision of proteomics analysis are high, and this often poses a challenge, especially when employing label-free quantification. One of the main contributors to the inaccuracy of label-free proteomics experiments is the variability of the instrumental response during LC-MS/MS runs. Such variability might include fluctuations in the electrospray current, transmission efficiency from the air-vacuum interface to the detector, and detection sensitivity. We have developed an in silico post-processing method of reducing these variations, and have thus significantly improved the precision of label-free proteomics analysis. For abundant blood plasma proteins, a coefficient of variation of approximately 1% was achieved, which allowed for sex differentiation in pooled samples and ≈90% accurate differentiation of individual samples by means of a single LC-MS/MS analysis. This method improves the precision of measurements and increases the accuracy of predictive models based on the measurements. The post-acquisition nature of the correction technique and its generality promise its widespread application in LC-MS/MS-based methods such as proteomics and metabolomics.