Multicenter Longitudinal Quality Assessment of MS-Based Proteomics in Plasma and Serum

Abstract

Advancing MS-based proteomics toward clinical applications evolves around developing standardized start-to-finish and fit-for-purpose workflows for clinical specimens. Steps along the method design involve the determination and optimization of several bioanalytical parameters such as selectivity, sensitivity, accuracy, and precision. In a joint effort, eight proteomics laboratories belonging to the MSCoreSys initiative including the CLINSPECT-M, MSTARS, DIASyM, and SMART-CARE consortia performed a longitudinal round-robin study to assess the analysis performance of plasma and serum as clinically relevant samples. A variety of LC-MS/MS setups including mass spectrometer models from ThermoFisher and Bruker as well as LC systems from ThermoFisher, Evosep, and Waters Corporation were used in this study. As key performance indicators, sensitivity, precision, and reproducibility were monitored over time. Protein identifications range between 300 and 400 IDs across different state-of-the-art MS instruments, with timsTOF Pro, Orbitrap Exploris 480, and Q Exactive HF-X being among the top performers. Overall, 71 proteins are reproducibly detectable in all setups in both serum and plasma samples, and 22 of these proteins are FDA-approved biomarkers, which are reproducibly quantified (CV < 20% with label-free quantification). In total, the round-robin study highlights a promising baseline for bringing MS-based measurements of serum and plasma samples closer to clinical utility.

Keywords: LC-MS/MS; R package mpwR; clinical specimen; longitudinal round-robin study; plasma; serum.

Figure 1

Study design (I) and descriptive summaries of the round-robin study data sets (II) including both plasma and serum analyses for Lab-LC-MS combination (A), LC-MS setups (B), LC systems (C), MS instruments (D), and acquisition mode (E) for one time point, respectively.

Figure 2

Protein IDs for plasma (A) and serum (B) for different setups. Measurements are color coded by time points. Median number of protein IDs per set up is shown as label.

Figure 3

Median number of protein identifications [abs.] and interquartile range (IQR) sorted in decreasing order based on median (a) and in increasing order based on IQR (b) on protein-level for plasma (I) and serum (II). Results are color coded by MS instrument. Labels show median number of protein identifications (a) or IQR (b).

Figure 4

Data completeness based on absolute numbers [abs.] of proteins for plasma (A) and serum (B) for different setups. Only full profiles, which refer to the presence of an identification in each technical replicate run per time point, are displayed and color coded by different time points. Mean and standard deviation as error bars are plotted in black.

Figure 5

Data completeness based on absolute numbers [abs.] on protein-level for plasma (A) and serum (B). Results of each time point are merged resulting in 15 runs per setup. Complete profiles refer to proteins present in all replicate runs (15), shared with at least 50% to be at least present in 50% of the replicate runs, sparse to be present in more than one run and less than 50% of the runs, and unique to be only present in one replicate run.

Figure 6

Quantitative precision coefficient of variation (CV) LFQ < 20% on absolute numbers [abs.] on protein group-level for plasma (A) and serum (B). The different time points are color coded.

Figure 7

Interlaboratory reproducibility - overlapping protein IDs [abs.] for plasma and serum with different levels of data completeness [%] based on all available data sets per sample type, respectively. For plasma 62 data sets and for serum 63 data sets are considered.

Figure 8

Quantitative precision across time points and per setup for FDA approved biomarker proteins in plasma. Red dashed horizontal line indicates CV 20%.

Figure 9

Quantitative precision across time points and per setup for FDA approved biomarker proteins in serum. Red dashed horizontal line indicates CV 20%.