Plasma Proteome Profiling Reveals the Dynamic Changes of Cholesterol Markers After Dual Immune Checkpoint Blockade Therapy

    Network-Based Elucidation of Colon Cancer Drug Resistance Mechanisms by Phosphoproteomic Time-Series Analysis

     Biomarker Identification and Risk Prediction Model Development for Lung Metastasis of Differentiated Thyroid Carcinoma Based on Primary Lesion Proteomics

    Optimizing Differential Expression Analysis for Proteomics Data via High-Performing Rules and Ensemble Inference

     Longitudinal Molecular Profiling Elucidates Immunometabolism Dynamics in Breast Cancer

     Temporal Dynamics of the Multi-omic Response to Endurance Exercise Training

     Molecular Pixelation: Single-Cell Spatial Proteomics

     Identifying Cell Types and Quantifying Cellular Spatial Relationships from Spatial Proteomics

Compared with monotherapy, dual blockade therapy (DBT) achieves better anti-tumour effects. Immunotherapy, however, has long lacked accepted biomarkers for predicting efficacy, which makes the search for suitable ones especially important. The team of Chen Ding at Fudan University and the team of You-Chao Jia at the Affiliated Hospital of Hebei University performed longitudinal plasma proteomic analysis of QL1706, a PD-1/CTLA-4 dual immune checkpoint blocker, and found that cholesterol metabolism was activated in the disease non-progression (DNP) group, together with the biomarker APOC3, which is highly correlated with activation of HDL particle remodelling in the DNP group. They proposed that PA, LDH and APOC3 may serve as potential predictive biomarkers of DBT efficacy. A machine learning model built on proteomic and clinical features achieved accurate prediction in the DBT cohort. Notably, the predictive power of this model also extends to anti-PD-1 treatment cohorts. At the same time, longitudinal exploration of these clinical and proteomic biomarkers indicates a durable response to DBT. This study broadens the biological understanding of plasma proteins in DBT and generates hypotheses that may serve as the basis for future clinical trials of precision immunotherapy response.

Network-Based Elucidation of Colon Cancer Drug Resistance Mechanisms by Phosphoproteomic Time-Series Analysis

Aberrant signalling pathway activity is a hallmark of tumour initiation and progression, and has guided the design of targeted inhibitors for more than 30 years. Yet adaptive resistance mechanisms induced by rapid, context-specific signalling networks compromise treatment efficacy. Drawing on advances in proteomics technology and network-based methodology, Andrea Califano of Columbia University and Yan-Sheng Liu of Yale University, working together, introduced virtual-enrichment-based signalling protein activity analysis (VESPA) to elucidate how cells respond to and adapt to drug perturbation, analysing phosphoproteomic changes in colorectal cancer cells treated with inhibitors. Kinase and phosphatase activities can be accurately inferred from the phosphorylation state of substrates, effectively accounting for signalling cross-talk and phosphorylation levels. This analysis elucidated the time-dependent signalling pathway responses to each drug perturbation; more importantly, CRISPR knockout experiments confirmed the adaptive responses and rewiring of the cells, indicating that the method can be widely applied to cancer and other diseases.

The rising global incidence of differentiated thyroid carcinoma (DTC) has led to a marked increase in DTC patients presenting with lung metastasis (LMDTC), a group that poses major challenges in clinical practice and creates an urgent need for effective risk stratification methods and lung metastasis prediction tools. Through proteomic analysis of a large number of primary lesion samples, with dual validation by parallel reaction monitoring and immunohistochemistry, researchers at Yunnan Cancer Hospital identified eight hub proteins as potential biomarkers. By expanding the sample size and analysing clinical features together with hub protein expression, three risk prediction models were built. Eight hub proteins — SUCLG1/2, DLAT, IDH3B, ACSF2, ACO2, CYCS and VDAC2 — were identified as potential biomarkers for predicting the risk of lung metastasis in DTC. The researchers developed and internally validated three risk prediction models that combine clinical features with hub protein expression. The study shows that the integrated prediction model had the best predictive performance, with the highest discrimination (AUC: 0.986) and calibration (Brier score: 0.043). Applying the integrated prediction model within a specific risk threshold (0–0.97) yielded the greatest clinical benefit. Finally, a nomogram was constructed on the basis of the integrated prediction model. As a large-sample study of metastatic DTC, this work identifies biomarkers through primary lesion proteomics and builds risk prediction models combining clinical features with hub protein biomarkers, providing important insights for predicting DTC lung metastasis and establishing personalized treatment strategies.

Optimizing Differential Expression Analysis for Proteomics Data via High-Performing Rules and Ensemble Inference

The identification of differentially expressed proteins in a proteomics workflow generally involves five key steps: raw data quantification, expression matrix construction, matrix normalization, missing value imputation (MVI) and differential expression analysis. Each step offers a large number of options, which makes it extremely challenging to determine the optimal workflow that maximizes the identification of differentially expressed proteins. To this end, Wilson Wen Bin Goh of Nanyang Technological University in Singapore and Professor Jin-Yan Li of the Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, conducted an extensive study on 24 gold-standard spike-in datasets involving 34,576 combined experiments. By applying frequent pattern mining to the top-ranked workflows, they discovered high-performing rules and demonstrated that optimality has a conserved property. Using machine learning, they confirmed that the optimal workflow is indeed predictable, with mean cross-validated F1 score and Matthew's correlation coefficient above 0.84. An ensemble inference method was introduced to integrate the results of individual best-performing workflows, thereby broadening the coverage of differential proteomes and resolving inconsistencies. Ensemble inference improved pAUC (by up to 4.61%) and G-mean (by up to 11.14%), and facilitated the effective pooling of information across different quantification methods (such as topN, directLFQ, MaxLFQ intensity and spectral counting). Further development and evaluation are needed to establish an accepted framework for ensemble inference across multiple proteomics workflows.

Although metabolic reprogramming in tumour cells and within the tumour microenvironment (TME) of breast cancer has been described in detail, little is known about the interplay between immune status and cancer metabolism during treatment. Researchers at the Karolinska Institute used proteomics, genomics and histopathology to characterize the immunometabolic profiles of tumour tissue samples collected longitudinally before, during and after neoadjuvant chemotherapy (NAC) in breast cancer patients. The study shows that dynamic changes in immune status, tumour metabolic proteins and metabolic phenotypes based on tumour cell gene expression profiles, measured before, during and after treatment, are associated with treatment response. Single-cell/single-nucleus RNA sequencing revealed distinct tumour and immune cell metabolic states between cold and hot tumours. Potential drivers of NAC identified through these analyses were validated in vitro. The study shows that the interaction between the tumour-intrinsic metabolic state and the TME is associated with treatment outcome, supporting the concept of immune modulation through targeting tumour metabolism.

Regular exercise benefits health and protects against disease, yet the underlying molecular mechanisms are poorly understood. Over eight weeks of endurance exercise training, the Molecular Transducers of Physical Activity Consortium analysed the temporal transcriptome, proteome, metabolome, lipidome, phosphoproteome, acetylproteome, ubiquitin proteome, epigenome and immunome in whole blood, plasma and 18 solid tissues of male and female rats. The resulting data compendium comprises 9,466 assays across 19 tissues, 25 molecular platforms and 4 training time points. Thousands of shared and tissue-specific molecular changes were identified, and sex differences were found across multiple tissues. Multi-omic and multi-tissue spatiotemporal analysis revealed a broad biological landscape of the adaptive response to endurance training, including extensive regulation of immune, metabolic, stress response and mitochondrial pathways. Many of the changes are relevant to human health, including non-alcoholic fatty liver disease, inflammatory bowel disease, cardiovascular health, and tissue injury and recovery. The data and analyses provided by this study will be a valuable resource for understanding and exploring the multi-tissue molecular effects of endurance training.

Molecular Pixelation: Single-Cell Spatial Proteomics

The spatial distribution of proteins on the cell surface governs important processes of the immune system, such as cell-to-cell communication and migration. Fluorescence microscopy or imaging flow cytometry is usually used to study the spatial distribution of proteins; these methods, however, have limitations in multiplexing and throughput, and their signal-to-noise ratio is affected by autofluorescence and colour spillover. Researchers at Pixelgen Technologies in Sweden, in collaboration with the Karolinska Institute, developed molecular pixelation (MPX), an optics-free technology that uses antibody–oligonucleotide conjugates (AOCs) and DNA-based molecular pixels to bind protein targets in fixed cells, detecting the arrangement of cell-surface proteins in a highly multiplexed manner. By adding unique molecular identifiers (UMIs), spatially proximal AOCs are sequentially linked into local neighbourhoods, enabling spatial analysis of protein arrangement; the resulting amplicons are then sequenced and protein spatial relationships are inferred from single-cell data graphs. This approach adds a spatial dimension to proteomics at the single-cell level and is expected to provide new insights into cell migration, cell activation, drug mode of action and drug target discovery.

Cellular composition and anatomical structure both have an important influence on the normal and abnormal function of organs. Spatial single-cell proteomics technologies such as imaging mass cytometry (IMC) and co-detection by indexing (CODEX) facilitate the study of cellular composition and structure by enabling high-throughput measurement of cells and their locations within intact tissue. The annotation of cell types and the quantification of their relative positions in tissue, however, remain challenging. To address the unmet needs of large-scale atlas datasets such as the Human Pancreas Analysis Program, Gregory W. Schwartz of the University of Toronto in Canada and Robert B. Faryabi of the University of Pennsylvania in the United States jointly developed AnnoSpat (annotator and spatial pattern finder), which uses neural networks and point process algorithms to automatically identify cell types and quantify cell-to-cell proximity relationships. Studies on IMC and CODEX data show that AnnoSpat performs better than other methods in annotating cell types quickly and accurately. In addition, applying AnnoSpat to cohorts of type 1 diabetes, non-diabetic autoantibody-positive and non-diabetic organ donors not only reproduced the known pathobiology of the islets, but also revealed dynamic changes in the number of pancreatic polypeptide cells in the islets and differential patterns of CD8+ T cell infiltration during the development of type 1 diabetes.