Brian C. Searle

INTRODUCTION: Reliable biomarkers for the diagnosis of chronic pancreatitis (CP) and pancreatic fibrosis severity are lacking, hindering effective treatment and management. Histologic fibrosis is a hallmark of late-stage CP, but noninvasive methods to evaluate fibrosis progression are limited. We used urine proteomics to discover biomarkers that identify patients with CP and predict fibrosis severity. METHODS: We performed a cross-sectional study of 130 total subjects (CP n = 50) selected based on clinical criteria in a tertiary care setting. Urine proteomics samples were quantified using data-independent acquisition mass spectrometry. Differential biomarker candidates were identified with false discovery rate-corrected pairwise comparisons. These proteins were validated with an independent paired urine and plasma sample cohort (n = 36). Machine learning was used to develop a protein panel that predicted Ammann scores for patients with histologic fibrosis. RESULTS: We found 34 proteins consistently differentially expressed between CP and controls in pairwise false discovery rate-controlled tests. Of these, 25 urine proteins outperformed 19 previously suggested CP blood-based biomarkers in an independent validation cohort. Isocitrate dehydrogenase (IDH1), calcyphosin (CAPS), synuclein gamma (SNCG), and protein S100-P (S100P) all produced receiver operator curve area under the curve values >0.95, while the best plasma marker was interleukin 2 receptor subunit alpha (receiver operator curve area under the curve = 0.80). A 12-protein panel of identified markers predicted fibrosis severity with a linear correlation R ² value of 0.61. DISCUSSION: We identified a panel of proteins that may diagnose CP in children and developed a model to predict pancreatic fibrosis severity, offering promising tools for improving diagnostics and patient care.

Recent developments in machine learning (ML) and deep learning have immense potential for applications in proteomics, such as generating spectral libraries, improving peptide identification, and optimizing targeted acquisition modes. Although new ML models are regularly published, the rate at which the community adopts these models is slow. This is in part due to a lack of findability and accessibility of these models as well as the technical challenges involved in incorporating these models into data analysis pipelines and demonstrating their reusability for end-users. Here we show Koina, an open-source decentralized and online-accessible model repository to facilitate publication of ML models. Koina enables ML model usage via an easy-to-use online interface, facilitating the integration of ML models in data analysis pipelines. Using the widely used FragPipe computational platform as an example, we demonstrate how Koina can be integrated with existing proteomics software tools and how these integrations improve data analysis.

Chronic infections and cancer cause T cell dysfunction known as exhaustion. This cell state is caused by persistent antigen exposure, suboptimal co-stimulation and a plethora of hostile factors that dampen protective immunity and limit the efficacy of immunotherapies^1,2,3,4. The mechanisms that underlie T cell exhaustion remain poorly understood. Here we analyse the proteome of CD8⁺ exhausted T (T_ex) cells across multiple states of exhaustion in the context of both chronic viral infections and cancer. We show that there is a non-stochastic pathway-specific discordance between mRNA and protein dynamics between T effector (T_eff) and T_ex cells. We identify a distinct proteotoxic stress response (PSR) in T_ex cells, which we term T_ex-PSR. Contrary to canonical stress responses that induce a reduction in protein synthesis^5,6, T_ex-PSR involves an increase in global translation activity and an upregulation of specialized chaperone proteins. T_ex-PSR is further characterized by the accumulation of protein aggregates and stress granules and an increase in autophagy-dominant protein catabolism. We establish that disruption of proteostasis alone can convert T_eff cells to T_ex cells, and we link T_ex-PSR mechanistically to persistent AKT signalling. Finally, disruption of T_ex-PSR-associated chaperones in CD8⁺ T cells improves cancer immunotherapy in preclinical models. Moreover, a high T_ex-PSR in T cells from patients with cancer confers poor responses to clinical immunotherapy. Collectively, our findings indicate that T_ex-PSR is a hallmark and a mechanistic driver of T cell exhaustion, which raises the possibility of targeting proteostasis pathways as an approach for cancer immunotherapy.

Orbitrap (OT)-based mass spectrometer platforms are a gold standard in high-resolution mass spectrometry, where their primary disadvantage is slower-scanning speed in comparison to time-of-flight or linear ion trap mass analyzers. In this study, we utilize long OT transients to extend the precursor dynamic range by modifying the selected ion monitoring method to multiplex several precursor m/z ranges into a single scan, which we call "multiple accumulation precursor mass spectrometry". Our approach requires no software or hardware modifications and hides the additional ion accumulation steps during the time it takes to make other Orbitrap measurements, producing precursor spectra with nearly 2× dynamic range and essentially no consequences. We collected data using both data-dependent acquisition (DDA) and data-independent acquisition (DIA) methods to evaluate a range of approaches. With DDA, MAP-MS precursor quantification improves with higher quality measurements. At the same time, DIA detection is enhanced by up to 11% when combining precursor and tandem mass spectra for peptide detection.

Ocean metaproteomics provides valuable insights into the structure and function of marine microbial communities. Yet, ocean samples are challenging due to their extensive biological diversity, which results in a very large number of peptides with a large dynamic range. This study characterized the capabilities of data independent acquisition (DIA) mode for use in ocean metaproteomic samples. Spectral libraries were constructed from discovered peptides and proteins using machine learning (ML) algorithms to remove the incorporation of false positives in the libraries. When compared with 1‐dimensional and 2‐dimensional data dependent acquisition analyses (DDA), DIA outperformed DDA both with and without gas phase fractionation. We found that larger discovered protein spectral libraries performed better, regardless of the geographic distance between where samples were collected for library generation and where the test samples were collected. Moreover, the spectral library containing all unique proteins present in the Ocean Protein Portal (OPP) outperformed smaller libraries generated from individual sampling campaigns. However, a spectral library constructed from all open reading frames (ORFs) in a metagenome was found to be too large to be workable, resulting in low peptide identifications due to challenges in maintaining a low false discovery rate with such a large database size. Given sufficient sequencing depth and validation studies, spectral libraries generated from previously discovered proteins can serve as a community resource, saving resequencing efforts. The spectral libraries generated in this study are available at the OPP to enable future ocean proteomic studies.

Mass spectrometry-based metaproteomics, the identification and quantification of thousands of proteins expressed by complex microbial communities, has become pivotal for unraveling functional interactions within microbiomes. However, metaproteomics data analysis encounters many challenges, including the search of tandem mass spectra against a protein sequence database using proteomics database search algorithms. We used a ground-truth dataset to assess a spectral library searching method against established database searching approaches. Mass spectrometry data collected by data-dependent acquisition (DDA-MS) was analyzed using database searching approaches (MaxQuant and FragPipe), as well as using Scribe with Prosit predicted spectral libraries. We used FASTA databases that included protein sequences from microbial species present in the ground-truth dataset along with background protein sequences, to estimate error rates and assess the effects on detection, peptide-spectral match quality, and quantification. Using the Scribe search engine resulted in more proteins detected at a 1% false discovery rate (FDR) compared to MaxQuant or FragPipe, while FragPipe detected more peptides verified by PepQuery. Scribe was able to detect more low-abundance proteins in the microbiome dataset and was more accurate in quantifying the microbial community composition. This research provides insights and guidance for metaproteomics researchers aiming to optimize results in their analysis of DDA-MS data.

Orbitrap (OT) -based mass spectrometer platforms are a gold standard in high-resolution mass spectrometry, where their primary disadvantage is slower-scanning speed in comparison to time-of-flight or linear ion trap mass analyzers. In this study, we utilize long OT transients to extend the precursor dynamic range by modifying the selected ion monitoring method to multiplex several precursor m/z ranges from 400 to 1000 m/z into a single scan called “ M ultiple A ccumulation P recursor M ass S pectrometry” (MAP-MS). Our approach requires no software or hardware modifications and hides the additional ion accumulation steps during the time it takes to make other Orbitrap measurements, producing precursor spectra with nearly 2× dynamic range and essentially no consequences. We collected data using both data-dependent acquisition (DDA) and data-independent acquisition (DIA) methods to evaluate a range of approaches. With DDA, MAP-MS precursor quantification improves with higher quality measurements. At the same time, DIA detection is enhanced by up to 11% when combining precursor and tandem mass spectra for peptide detection.

Advances in proteomics and mass spectrometry enable the study of limited cell populations, where high-mass accuracy instruments are typically required. While triple quadrupoles offer fast and sensitive low-mass specificity measurements, these instruments are effectively restricted to targeted proteomics. Linear ion traps (LITs) offer a versatile, cost-effective alternative capable of both targeted and global proteomics. Here, we describe a workflow using a hybrid quadrupole-LIT instrument that rapidly develops targeted proteomics assays from global data-independent acquisition (DIA) measurements without high-mass accuracy. Using an automated software approach for scheduling parallel reaction monitoring assays (PRM), we show consistent quantification across three orders of magnitude in a matched-matrix background. We demonstrate measuring low-level proteins such as transcription factors and cytokines with quantitative linearity below two orders of magnitude in a 1 ng background proteome without requiring stable isotope-labeled standards. From a 1 ng sample, we found clear consistency between proteins in subsets of CD4⁺ and CD8⁺ T cells measured using high dimensional flow cytometry and LIT-based proteomics. Based on these results, we believe hybrid quadrupole-LIT instruments represent a valuable solution to expanding mass spectrometry in a wide variety of laboratory settings.

Mass spectrometry (MS)‐based metaproteomics is used to identify and quantify proteins in microbiome samples, with the frequently used methodology being data‐dependent acquisition mass spectrometry (DDA‐MS). However, DDA‐MS is limited in its ability to reproducibly identify and quantify lower abundant peptides and proteins. To address DDA‐MS deficiencies, proteomics researchers have started using Data‐independent acquisition mass spectrometry (DIA‐MS) for reproducible detection and quantification of peptides and proteins. We sought to evaluate the reproducibility and accuracy of DIA‐MS metaproteomic measurements relative to DDA‐MS using a mock community of known taxonomic composition. Artificial microbial communities of known composition were analyzed independently in three laboratories using DDA‐ and DIA‐MS acquisition methods. In this study, DIA‐MS yielded more protein and peptide identifications than DDA‐MS in each laboratory for the particular instruments and software parameters chosen. In addition, the protein and peptide identifications were more reproducible in all laboratories and provided an accurate quantification of proteins and taxonomic groups in the samples. We also identified some limitations of current DIA tools when applied to metaproteomic data, highlighting specific needs to improve DIA tools enabling analysis of metaproteomic datasets from complex microbiomes. Ultimately, DIA‐MS represents a promising strategy for MS‐based metaproteomics due to its large number of detected proteins and peptides, reproducibility, deep sequencing capabilities, and accurate quantitation.