Research On Performance Evaluation Of Mass Spectrometer Based On Quantitative Proteomics Data

With the innovation of technology for mass spectrometry and widespread application of Liquid chromatography-tandem mass spectrometry(LC-MS/MS),largescale proteomic studies in the field of life science can be readily carried out to serve the researches of drug discovery and precision medicine.Reproducibility is a crucial foundation of scientific research,which requires higher standards for the quality of data output in the process of large-scale research.MS-based proteomics technologies allow for identifying and quantifying thousands of proteins in complex sample but the consistent and accurate quantification of proteins depends on the performance of instruments,acquisition methods and data analysis software and so on.In the field of proteomics,the relevant work with quality control has explored the effects from different source of variability,including sample preparation,LC conditions,and bioinformatics analysis methods on results,but not yet considering LC-MS/MS performance.In fact,the real-time evaluation for instrument performance is vital to maintain the smooth running of mass spectrometers during project progress.Hence,we collected 1,200 quantitative proteomics datasets from homemade 293 T standards runs for the entire year of 2016,produced by 10 Orbitrap-based instruments,in National Center for Protein Sciences · Beijing.Before data analysis,these experiments with incomplete or missing experimental records were excluded on basis of the principle of data checking,then 720 experiments produced by 6 instruments were selected for follow-up study.Combing with manual inspection and bioinformatics analysis,a series of quality control steps including spectrum review,RT review for 8 peaks,normalization,PCA,clustering and outlier analysis were performed to determine 458 benchmark experiments from 6 instruments,respectively.The analysis based on benchmarking sets indicated that our LC-MS platforms had a higher qualitative reproducibility and quantitative consistency.At peptide and protein level,the peptide/proteins lists from pairs of technical replicates overlapped by around 45% and 73%,the intra-instrument coefficient of variations(CVs)were 25% and 33%.This result objectively represented the overall performance of Orbitrap-based MS for peptide and protein identification in complex samples,and kept consistence with previous studies,where the repeatability and reproducibility of proteins were higher than those of peptides.In order to explore the sensitivity of differently expressed proteins detected by various instruments,the results from “Quartet” standards data analysis showed differently expressed proteins identified by diverse LC-MS/MS were dfferent but the enrichment pathways using those proteins were highly consistent,which reflected intrinsic physiological processes of cell used for standards preparation.Given that the undulation of instrument performance can give rise to perturbation for quantitative response,a reference peptides set for each instrument was generated to construct the scoring model,which was utilized for monitoring instrument performance.Based on manual inspection and data analysis,the low-quality experiment sets were labeled as negative while the high-quality experiment sets were labeled as positive,receiver operating characteristic(ROC)curve was applied to determine the cutoff and characterized the critical state of instrument performance.The average area under the curve(AUC)was 0.98,showing reliability of above-mentioned scoring models.The results from an independent validation set demonstrated that the model based on quantitative reference peptides set could help operators objectively evaluate status of instruments,timely perform imperative maintenance for instruments with poor performance.Based on the results of the reproducibility assessment above,we further analyzed the factors affecting the reproducibility in detail.Combining the proteomics solution in this paper,we put forward some measures to improve the reproducibility of the results,which consisted of sample preparation,LC-MS/MS parameter setting and bioinformatics method application.Additionally,our study also expands to quality control work in the field of phosphoproteomics.We developed an R package,called PhosMap,aiming to help most biological researchers with few programming skills perform quality control on their phosphoproteomics data and accelerate their research progress.