Application of clinical assay quality control (QC) to multivariate proteomics data: a workflow exemplified by 2-DE QC.

Proteomics technologies can be effective for the discovery and assay of protein forms altered with disease. However, few examples of successful biomarker discovery yet exist. Critical to addressing this is the widespread implementation of appropriate QC (quality control) methodology. Such QC should combine the rigour of clinical laboratory assays with a suitable treatment of the complexity of the proteome by targeting separate assignable causes of variation. We demonstrate an approach, metric and example workflow for users to develop such targeted QC rules systematically and objectively, using a publicly available plasma DIGE data set. Hierarchical clustering analysis of standard channels is first used to discover correlated groups of features corresponding to specific assignable sources of technical variation. These effects are then quantified using a statistical distance metric, and followed on control charts. This allows measurement of process drift and the detection of runs that outlie for any given effect. A known technical issue on originally rejected gels was detected validating this approach, and relevant novel effects were also detected and classified effectively. Our approach was effective for 2-DE QC. Whilst we demonstrated this in a retrospective DIGE experiment, the principles would apply to ongoing QC and other proteomic technologies. BIOLOGICAL SIGNIFICANCE: This work asserts that properly carried out QC is essential to proteomics discovery experiments. Its significance is that it provides one possible novel framework for applying such methods, with a particular consideration of how to handle the complexity of the proteome. It not only focusses on 2DE-based methodology but also demonstrates general principles. A combination of results and discussion based upon a publicly available data set is used to illustrate the approach and allows a structured discussion of factors that experimenters may wish to bear in mind in other situations. The demonstration is on retrospective data only for reasons of scope, but the principles applied are also important for ongoing QC, and this work serves as a step towards a later demonstration of that application. This article is part of a Special Issue entitled: Standardization and Quality Control in Proteomics.

An introduction to statistical process control in research proteomics.

BACKGROUND: Statistical process control is a well-established and respected method which provides a general purpose, and consistent framework for monitoring and improving the quality of a process. It is routinely used in many industries where the quality of final products is critical and is often required in clinical diagnostic laboratories [1,2]. To date, the methodology has been little utilised in research proteomics. It has been shown to be capable of delivering quantitative QC procedures for qualitative clinical assays [3] making it an ideal methodology to apply to this area of biological research. OBJECTIVE: To introduce statistical process control as an objective strategy for quality control and show how it could be used to benefit proteomics researchers and enhance the quality of the results they generate. RESULTS: We demonstrate that rules which provide basic quality control are easy to derive and implement and could have a major impact on data quality for many studies. CONCLUSIONS: Statistical process control is a powerful tool for investigating and improving proteomics research work-flows. The process of characterising measurement systems and defining control rules forces the exploration of key questions that can lead to significant improvements in performance. BIOLOGICAL SIGNIFICANCE: This work asserts that QC is essential to proteomics discovery experiments. Every experimenter must know the current capabilities of their measurement system and have an objective means for tracking and ensuring that performance. Proteomic analysis work-flows are complicated and multi-variate. QC is critical for clinical chemistry measurements and huge strides have been made in ensuring the quality and validity of results in clinical biochemistry labs. This work introduces some of these QC concepts and works to bridge their use from single analyte QC to applications in multi-analyte systems. This article is part of a Special Issue entitled: Standardization and Quality Control in Proteomics.

Considerations for powering a clinical proteomics study: Normal variability in the human plasma proteome.

Proteomics is increasingly being applied to the human plasma proteome to identify biomarkers of disease for use in non-invasive assays. 2-D DIGE, simultaneously analysing thousands of protein spots quantitatively and maintaining protein isoform information, is one technique adopted. Sufficient numbers of samples must be analysed to achieve statistical power; however, few reported studies have analysed inherent variability in the plasma proteome by 2-D DIGE to allow power calculations. This study analysed plasma from 60 healthy volunteers by 2-D DIGE. Two samples were taken, 7 days apart, allowing estimation of sensitivity of detection of differences in spot intensity between two groups using either a longitudinal (paired) or non-paired design. Parameters for differences were: two-fold normalised volume change, α of 0.05 and power of 0.8. Using groups of 20 samples, alterations in 1742 spots could be detected with longitudinal sampling, and in 1206 between non-paired groups. Interbatch gel variability was small relative to the detection parameters, indicating robustness and reproducibility of 2-D DIGE for analysing large sample sets. In summary, 20 samples can allow detection of a large number of proteomic alterations by 2-D DIGE in human plasma, the sensitivity of detecting differences was greatly improved by longitudinal sampling and the technology was robust across batches.