Reproducible research is the bedrock of experimental science. To enable the deployment of large-scale proteomics, we assess the reproducibility of mass spectrometry (MS) over time and across instruments and develop computational methods for improving quantitative accuracy. We perform 1560 data independent acquisition (DIA)-MS runs of eight samples containing known proportions of ovarian and prostate cancer tissue and yeast, or control HEK293T cells. Replicates are run on six mass spectrometers operating continuously with varying maintenance schedules over four months, interspersed with~5000 other runs. We utilise negative controls and replicates to remove unwanted variation and enhance biological signal, outperforming existing methods. We also design a method for reducing missing values. Integrating these computational modules into a pipeline (ProNorM), we mitigate variation among instruments over time and accurately predict tissue proportions. We demonstrate how to improve the quantitative analysis of large-scale DIA-MS data, providing a pathway toward clinical proteomics.

No abstract

Background: One of the most significant challenges in colorectal cancer (CRC) management is the use of compliant early stage population-based diagnostic tests as adjuncts to confirmatory colonoscopy. Despite the near curative nature of early clinical stage surgical resection, mortality remains unacceptably high-as the majority of patients diagnosed by faecal haemoglobin followed by colonoscopy occur at latter stages. Additionally, current populationbased screens reliant on fecal occult blood test (FOBT) have low compliance (~ 40%) and tests suffer low sensitivities. Therefore, blood-based diagnostic tests offer survival benefits from their higher compliance (≥ 97%), if they can at least match the sensitivity and specificity of FOBTs. However, discovery of low abundance plasma biomarkers is difficult due to occupancy of a high percentage of proteomic discovery space by many high abundance plasma proteins (e.g., human serum albumin). Methods: A combination of high abundance protein ultradepletion (e.g., MARS-14 and an in-house IgY depletion columns) strategies, extensive peptide fractionation methods (SCX, SAX, High pH and SEC) and SWATH-MS were utilized to uncover protein biomarkers from a cohort of 100 plasma samples (i.e., pools of 20 healthy and 20 stages I-IV CRC plasmas). The differentially expressed proteins were analyzed using ANOVA and pairwise t-tests (p < 0.05; fold-change > 1.5), and further examined with a neural network classification method using in silico augmented 5000 patient datasets. Results: Ultradepletion combined with peptide fractionation allowed for the identification of a total of 513 plasma proteins, 8 of which had not been previously reported in human plasma (based on PeptideAtlas database). SWATH-MS analysis revealed 37 protein biomarker candidates that exhibited differential expression across CRC stages compared to healthy controls. Of those, 7 candidates (CST3, GPX3, CFD, MRC1, COMP, PON1 and ADAMDEC1) were validated using Western blotting and/or ELISA. The neural network classification narrowed down candidate biomarkers to 5 proteins (SAA2, APCS, APOA4, F2 and AMBP) that had maintained accuracy which could discern early (I/II) from late (III/IV) stage CRC. Conclusion: MS-based proteomics in combination with ultradepletion strategies have an immense potential of identifying diagnostic protein biosignature.