On February 26, 2025, Tiannan Guo from Westlake University School of Medicine, Judith A. Steen from Harvard Medical School, and Matthias Mann from the Max Planck Institute of Biochemistry published a review titled "Mass-spectrometry-based proteomics: from single cells to clinical applications" in Nature.

This is a summary of the field nearly a decade after the review published in 2016 by proteomics two leading figures Ruedi Aebersold and Matthias Mann.

2025 review article

2016 review article

 

This 2025 article systematically reviews the development and latest progress of mass spectrometry-based proteomics technology in the past decade and details the overall process innovation in this field.

The authors emphasize that through technological innovation and promotion from hardware to software, mass spectrometry-based proteomics has achieved large-scale detection from conventional biological samples to single cells and even spatial resolution analysis significantly improving the depth and accuracy of protein identification, quantification, and structural analysis.

The article also discusses the enormous potential and challenges of translating proteomics to clinical applications including the discovery of disease biomarkers and the development of multi-marker detection methods, and further points out and emphasizes the application prospects of targeted proteomics technology in disease diagnosis and precision treatment.

At the end of the article, the authors look forward to the future development direction of proteomics, believing that with the promotion of automation, multi-omics data integration, especially artificial intelligence this field will further break through existing bottlenecks, providing new perspectives and pathways for revealing complex biological systems and disease mechanisms, and ultimately promoting the realization of precision medicine.

 

1. Overview of mass spectrometry-based proteomics technology

In recent years, with the introduction of automated and standardized technologies the efficiency and reproducibility of biological sample pretreatment processes have been significantly improved. A variety of sample types, including formalin-fixed paraffin-embedded (FFPE) tissues and archaeological samples, can be successfully analyzed.

In terms of chromatographic separation, thanks to the high sensitivity of modern mass spectrometers, the traditional nanoliter flow rate has been gradually replaced by the microliter flow rate, greatly improving the robustness of chromatographic methods. At the same time, mass spectrometer instruments themselves are constantly improving. Although Orbitrap analyzers still dominate, time-of-flight (TOF) analyzers have been revived. For example, timsTOF uses parallel accumulation—serial fragmentation (PASEF) technology to improve peptide sequencing efficiency through ion mobility separation, while the Astral analyzer uses multi-reflection TOF combined with Orbitrap for MS1 analysis, achieving almost lossless ion detection and excellent resolution.

In addition, data acquisition strategies have also undergone significant changes. The traditional data-dependent acquisition (DDA) mode is limited by the bias and randomness of signal detection, and more and more research uses data-independent acquisition (DIA) mode.

DIA scans and fragments peptide ions in a continuous window mode across the full mass range, allowing unbiased detection of essentially all peptide ions, greatly reducing missing values between different samples. Combined with advanced machine learning and deep learning algorithms, it can significantly improve the efficiency of interpreting complex DIA MS spectra. For the quantitative needs of clinical applications, targeted proteomics (S/MRM) technology is rapidly developing, achieving high-throughput, absolute quantification of target peptides through short chromatographic gradients.

Mass spectrometry-based proteomics workflow

 

The article particularly emphasizes artificial intelligence the transformative role in mass spectrometry-based proteomics research. Deep learning algorithms can not only predict the performance of peptides throughout the entire process (such as peptide retention time and specificity), improving the accuracy of identification and quantification, but also show great potential in biomarker mining and interpretation of biological significance. The application of integrating large language models (such as ChatGPT, DeepSeek, etc.) in spectral analysis is expected to accelerate data interpretation and hypothesis generation in the future, pushing proteomic technology to new heights. 。

Overall, these technological innovations have made mass spectrometry-based proteomics a high-throughput, high-sensitivity, and highly applicable analytical platform, enabling wide-ranging applications from single-cell research to large-scale clinical trials.

 

2. Applications of mass spectrometry proteomics

The article demonstrates the multi-dimensional applications of mass spectrometry-based proteomics in analyzing biological systems from seven different perspectives.

Through expression proteomics, we can understand the quantitative changes of proteins under different conditions; interaction proteomics reveals the complex molecular interconnected network within cells; post-translational modifications (PTMs) and structural proteomics provide detailed descriptions of the dynamic changes of proteins from the aspects of functional regulation and structural analysis, respectively; chemoproteomics and single-cell proteomics further expand the application boundaries, the former provides new tools for drug target discovery and mechanism research, and the latter breaks through the technical bottlenecks of cell heterogeneity research; while spatial proteomics maintains the original spatial information of the tissue, allowing us to understand biological functions and pathological processes more intuitively from the tissue and subcellular levels.

(1) Expression proteomics

Expression proteomics mainly focuses on the changes in protein abundance under different physiological or pathological conditions. Through large-scale quantitative analysis, researchers can reveal how cells respond to various risk factors such as lifestyle, gene mutations, environmental toxins, and infections, thus providing a basis for drug development and companion diagnostics. Current technology can simultaneously quantify tens of thousands of proteins in common cell lines, and through pre-fractionation technology, it can even achieve near-comprehensive coverage. This not only demonstrates the differences in regulatory mechanisms between the proteome and the transcriptome, but also emphasizes the importance of proteins directly reflecting cellular function.

Recommended related reading:

Literature Review | Cell bombshell: Fudan team comprehensively maps the human health and disease proteome atlas [Collection];

Literature Review | CRISPR-Cas9 gene-edited HAP1 cell model assists in the research of cancer and other complex diseases [Treasure].

(2) Interaction proteomics

This field focuses on resolving the interactions between proteins and other molecules (such as proteins, RNA, DNA, small molecule drugs). Traditional affinity purification methods combined with emerging proximity labeling techniques allow researchers to construct nearly complete protein interaction networks, providing strong data support for understanding complex intracellular signaling and regulatory mechanisms.

Meanwhile, using gene editing tools and AI assistance (such as AlphaFold-Multimer), further analysis of the structure and function of multi-protein complexes provides a new perspective for the study of molecular mechanisms in pathological states.

Recommended related reading:

Technical Special | Protein proximity labeling (PL): A new standard configuration for protein interaction research [Collection]

(3) PTMs (Post-translational modification proteomics)

Post-translational modifications play a crucial role in regulating protein function, localization, and interactions. Through automated enrichment, ion mobility separation, and other technologies, researchers can now accurately detect and quantify tens of thousands of PTM sites, such as phosphorylation, ubiquitination, and acetylation.

Such high-resolution analysis not only reveals the dynamic changes of signaling pathways in processes such as the cell cycle, signal transduction, and aging, but also uses machine learning and other AI technologies to help screen for functional modifications, thereby deepening our understanding of the complexity of cellular regulatory networks.

(4) Structural proteomics

Structural proteomics uses MS technology to study the three-dimensional structure and dynamic changes of proteins, providing information that is difficult to obtain through traditional structural biology. Methods include hydrogen-deuterium exchange (HDX-MS), cross-linking mass spectrometry (XL-MS), and limited proteolysis, which can reveal protein conformational changes, intra-protein complex interactions, and binding site information on a large scale.

These technologies provide structural support for resolving protein function and lay the foundation for applications such as drug design and antibody development.

(5) Chemoproteomics

Chemoproteomics focuses on systematically detecting the interaction between proteins and small molecules, becoming an important tool for drug discovery and target identification . The article mentions that activity-based protein profiling (ABPP) uses chemical probes to directly capture the enzymatic activity of specific protein families, while thermal proteome profiling (TPP) identifies the direct targets of small molecule drugs through temperature-induced protein denaturation.

Recommended related reading:

Literature Review | 2024 Trends in Innovative Drug Development in China: Breakthroughs and Prospects for New Drug Targets

In recent years, by screening degraders, researchers have even been able to design new strategies that specifically promote protein degradation for "undruggable" targets, greatly expanding the scope of therapeutic intervention.

Recommended related reading:

Particle Man Collaboration I TRIM56 mediates the production of type I interferon to inhibit the replication of rickettsial pathogens

(6) Single-cell proteomics

Single-cell proteomics aims to solve the problem that traditional proteomics cannot capture intercellular heterogeneity. Due to the extremely low absolute amount of protein in a single cell (only fifty to five hundred picograms), this field faces extremely high sensitivity requirements.

To this end, researchers have developed a series of innovative methods, including nanoscale droplet processing technology, the SCoPE method using stable isotope labeling (isobaric labeling) to achieve single-cell and carrier channel binding, and microfluidic technology, and combined with DIA data acquisition strategies, effectively improving the accuracy of single-cell protein quantification. With these technologies, it is now possible to identify up to 5000 different proteins in a single tumor cell, providing more direct and accurate data support for revealing cell state and function.

(7) Spatial proteomics

Although single-cell analysis reveals the intrinsic heterogeneity of cells, it often lacks spatial information at the tissue level. Spatial proteomics retains the spatial distribution of proteins in the sample, combines high-resolution imaging, AI-driven cell classification, and micro-dissection techniques to successfully construct protein expression profiles of various cell types or states in tissues. This method can not only reveal intercellular physical interactions, but also provide new molecular evidence for understanding tissue structure and pathological states (such as applications in drug-induced skin lesions), promoting the development of precision medicine.

Recently, this field has seen deep visual proteomics (DVP) and expansion proteomics and other technologies.

Single-cell and spatial proteomics

Overall, these seven directions complement each other, jointly promoting the continuous advancement of proteomics technology, providing rich and precise data support for basic biological research and clinical applications.

 

3. Applying proteomics to the clinic

The article also explores how to translate mature proteomics technologies from laboratory research into clinical applications to improve disease diagnosis, prognosis assessment, and treatment decisions.

Blood samples have always been considered a treasure trove of potential biomarkers. However, the abundance of proteins in plasma spans a huge range (more than ten orders of magnitude), posing a significant challenge to quantitative analysis.

With advancements in automated liquid handling systems, sophisticated HPLC (high-performance liquid chromatography) systems, and highly sensitive mass spectrometers (such as timsTOF and Astral), existing mass spectrometry workflows can now effectively analyze untreated plasma samples. Through partial process optimization, hundreds of samples can be rapidly analyzed daily, facilitating research on large-scale clinical cohorts.

For example, a workflow using a liquid handler, Evosep HPLC system, and timsTOF mass spectrometer successfully and rapidly analyzed 1,612 samples, playing a key role in the discovery of sepsis biomarkers. Using the Astral mass spectrometer, this process can detect approximately 1,000 protein types in a single injection, processing hundreds of samples daily and providing robust data support for clinical cohort analysis.

Meanwhile, researchers are working to further improve the detection depth of plasma proteins through novel non-specific enrichment strategies (such as enrichment schemes using nanoparticles), aiming for quantitative analysis of thousands of proteins to more comprehensively explore potential biomarkers.

The article also discusses the potential of other body fluids (such as urine and tears) in clinical applications. Body fluid samples are gaining attention due to their relatively less invasive collection methods and high correlation with specific diseases.

Simultaneously, proteomic analysis of tissue samples is also advancing. Projects like CPTAC, through multi-level analysis of numerous tumor samples, have not only revealed tumor subtypes and driver mutations but also provided a molecular basis for future personalized treatment.

Although mass spectrometry shows great potential in biomarker discovery, the translation from disease biomarker discovery to clinical testing faces many challenges, such as standardized technical processes, the cost of mass spectrometry-based detection methods, and stringent clinical validation requirements (including both policy and technical aspects).

Therefore, the article emphasizes the need for more targeted proteomics strategies in the future, combined with internal standard quantification and standardized processes, to ensure the accuracy and reproducibility of detection results, thereby meeting the requirements of clinical and regulatory authorities.

 

4. Future Outlook of Proteomics and the Role of AI

Finally, the article looks forward to the future development prospects of mass spectrometry-based proteomics and focuses on the core role of AI in this field.

With continuous advancements in sample preparation, chromatographic separation, mass spectrometry instrumentation, and data acquisition strategies, proteomics will achieve deeper protein coverage, extracting richer information from minute samples. Simultaneously, automated and standardized workflows will make this technology more accessible, benefiting a wider range of research communities. Although the high-throughput characteristics of the latest generation of mass spectrometers partially offset their high cost, reducing equipment and operating costs remains a significant bottleneck for further promotion.

At the same time, AI is driving innovation in proteomics at multiple levels. Deep learning algorithms can not only optimize data acquisition and real-time decision-making but also show great potential in peptide and protein identification, quantification, and structural analysis, helping to extract key biological information from massive datasets.

AI-enabled proteomics. AI applications range from experimental design and sample preparation to data interpretation, prediction, and integration of different omics data, aiming to improve the efficiency and depth of proteomics research.

 

The article further predicts that with the accumulation of more proteomics data (especially perturbation proteomics data ), researchers will be able to build "foundation models" with a deep understanding of biology. These models may even be able to automatically translate from spectra to amino acid sequences, giving proteomics stronger analytical and predictive capabilities.

Recommended Related Reading: Strategic Cooperation | Leman Bio Wins Bid for Xihub Laboratory KO Cell Lysis Liquid Technical Service Project

 

In addition, the integration of multi-omics data (including genomics, transcriptomics, and metabolomics) will play a key role in revealing intracellular molecular networks and complex disease mechanisms. The establishment of open-source software, shared data repositories, and standardized experimental and data analysis processes will also promote global cooperation and knowledge dissemination.

The article paints a promising future picture: driven by technological innovation, computational methods, and interdisciplinary collaboration, proteomics will continuously break through existing limitations, bringing revolutionary changes to basic biological research and precision medicine.

Literature Link: https://www.nature.com/articles/s41586-025-08584-0

Note: The specific content is subject to the English literature. The Chinese interpretation is for reference only. Also, due to space limitations, this article cannot provide a complete summary and citation of all progress in the past decade. We apologize for any inconvenience!