Quality Control for high-throughput Quantitative Proteomics - Harnessing the potential of label-free LC-MS

Abstract: Multiplex protein quantification, proteomics, is essential for uncovering new biomarkers and understanding biology. Liquid chromatography coupled to mass spectrometry (LC-MS) is the predominant technique for these measurements. In this thesis, optimization and quality control strategies for different phases of the LC-MS pipeline have been developed. Due to the inherent complexity and dynamic range of the proteome, pre-fractionation is commonly employed in a general LC-MS setting to uncover low-abundant proteins. In paper I, we performed a (qualitative) comparative study of the proteins identified in a shotgun setup. It was found that SDS-PAGE, coupled to tryptic digestion, had the largest yield. In paper II, we introduced a quantitative quality control method on peptide level for the relatively computationally demanding, but experimentally high-throughput, label-free LC-MS workflow, where we discovered both considerable differences as well as complementary properties between software solutions. The two integral parts of label-free data processing, feature detection (peptide quantification) and alignment (peptide identity propagation between samples), were further examined in papers III and IV. The complementary quality of software modules discovered in paper II was utilized in paper III, where we showed that a combination of different feature detection methods results in higher peptide coverage beneficial for downstream statistical inference. Since the establishment of proteomics as a high-throughput science, large-scale bioinformatics has become necessity. In paper IV, an alignment algorithm was developed where parameters are estimated on the fly from underlying data, an important step towards avoiding turning data processing into the future bottleneck of proteomics. Finally, in paper V, we performed an extensive evaluation of the relative potential of discovery and validation LC-MS (shotgun and SRM), based on the method introduced in paper II. Extensive data handling optimization was performed from normalization to quantification and statistical implications were assessed. In summary, we have shown that large-scale label-free LC-MS can be combined with equally high-throughput quality control to create a competitive option in the hunt for biomarkers.