Introduction

Measurement of plasma in large cohorts enables a variety of studies including biomarker discovery, disease progression tracking and genomic variant analysis. Mass spectrometry (MS) is a powerful technology for unbiased and species-agnostic protein identification. However, challenges for population scale MS measurement have persisted due to dynamic range of plasma protein abundance, multiple sources of analytical variability, and limitations in conventional computational frameworks. Here we demonstrate a sensitive, robust, and efficient integrated workflow for MS proteomics measurement at population scale.

Methods

In a blinded study of 1,600 samples, we leveraged the Proteograph™ XT product suite to combine nanoparticle enrichment with plate-based automated sample preparation, MS measurement and cloud-scalable search and processing resources. MS measurements were made using the Orbitrap™ Astral™ MS in data-independent acquisition mode at a throughput of 40 samples per day with robust microflow chromatographic separation. Plate design included digestion, enrichment, cleanup, and MS controls to monitor independent components of platform performance. An additional pooled sample was deposited on every plate to characterize technical reproducibility of the workflow. Search of raw data files was enabled by a novel cloud-based implementation of DIA-NN version 1.8.1 match-between runs (MBR), titled scalable MBR, available in the Proteograph Analysis Suite (PAS) software.

Results

In total, approximately 8,100 protein groups were represented in this cohort by measurement of more than 104,000 peptide sequences. Plate design enabled continuous quality control monitoring to support the importance of automated sample preparations. For each control, monitoring metrics included depth, quantitative robustness, outlier analysis, correlation analysis, and dimensional reduction sample representation. Across the study, technical reproducibility of preparation and MS was observed at approximately 30% precursor-level coefficient of variation, significantly below non-technical replicate variance. Using a novel metric, we were able to infer interpretable biological signal without prior knowledge of study design. Additionally, the implementation of scalable MBR in PAS enabled completion of search in under one day, with projected completion more than twenty-times faster than equivalent search on a high-performance workstation.

Conclusions

The integration of automated enrichment and sample preparation, robust MS configuration and cloud informatics demonstrates the capacity of Proteograph XT to effectively scale to population-level studies. Furthermore, our on-plate controls validate the ability of this platform to generate reliable and reproducible measurements without laborious protocols or difficult to implement algorithmic pipelines.