Application of FABP3 in prediction of irAEs

Abstract

The invention discloses an application of FABP3 in prediction of irAEs. The FABP3 can well predict the occurrence and severity of the irAEs through the discovery queue, the prediction performance of the FABP3 is further verified through the verification queue, a risk prediction model for predicting the irAEs is constructed, the research of the application is expected to improve the management of the irAEs, and a more personalized, effective and safe immunotherapy scheme is provided for cancer patients.

Description

Technical Field

[0001] The present invention belongs to the field of biomedicine, and specifically relates to the application of FABP3 in predicting irAEs. Background Art

[0002] Anti-programmed death 1 (PD-1) / anti-programmed death ligand 1 (PD-L1) antibody therapy activates the body's immune response and disrupts immune homeostasis. The resulting immune-related adverse events (irAEs) can affect organs throughout the body and may lead to death in severe cases.

[0003] The occurrence of immune-related adverse reactions is highly uncertain and often unpredictable, and can occur within weeks to months after the start of immunotherapy. The occurrence of irAEs is the result of overactivation of the immune system, which can lead to immune-mediated damage to normal tissues and organs, and its manifestations range from mild transient symptoms to severe life-threatening conditions. The nature of irAEs is difficult to predict and the severity varies, so reliable biomarkers are urgently needed for non-invasive monitoring, early diagnosis, and effective management of these adverse events. Summary of the invention

[0004] To make up for the deficiencies of the prior art, the present invention provides the application of FABP3 in predicting irAEs.

[0005] To achieve the above purpose, the present invention adopts the following technical solution.

[0006] A first aspect of the present invention provides a product for predicting the risk of irAEs / predicting the severity of irAEs, wherein the product comprises a reagent for detecting the expression level of FABP3.

[0007] Furthermore, the product includes a chip, a kit or a nucleic acid membrane strip.

[0008] Furthermore, the kit also includes a container.

[0009] Furthermore, the kit also includes instructions.

[0010] Furthermore, the kit further comprises a buffer.

[0011] Furthermore, the kit also includes packaging materials.

[0012] A second aspect of the present invention provides a model for predicting the risk of irAEs / predicting the severity of irAEs, wherein the model includes FABP3.

[0013] Furthermore, the model also includes IL1RL1.

[0014] The third aspect of the present invention provides the use of a reagent for detecting FABP3 expression levels in the preparation of a product for predicting the risk of irAEs / predicting the severity of irAEs.

[0015] Further, the reagent is selected from any one of the following groups:

[0016] (1) an oligonucleotide probe that specifically recognizes the FABP3 gene;

[0017] (2) Primers that specifically amplify the FABP3 gene;

[0018] (3) A binding agent that specifically binds to the protein encoded by the FABP3 gene.

[0019] Furthermore, the product also includes a reagent for detecting the expression level of IL1RL1.

[0020] Furthermore, the reagent also includes a detectable marker.

[0021] Further, the detectable markers include UV-Vis labels, near-infrared labels, fluorescent groups, phosphorescent groups, magnetic spin resonance labels, photosensitizers, photocleavable moieties, chelating centers, heavy atoms, radioisotopes, isotope detectable spin resonance labels, paramagnetic moieties, chromophores, and luminophores.

[0022] The fourth aspect of the present invention provides the application of FABP3 in constructing a model for predicting the risk of irAEs / predicting the severity of irAEs.

[0023] The fifth aspect of the present invention provides the application of FABP3 in constructing a system for predicting the risk of irAEs / predicting the severity of irAEs.

[0024] A sixth aspect of the present invention provides a system for predicting the risk of irAEs / predicting the severity of irAEs, the system comprising:

[0025] Acquisition unit: used for acquiring the expression level of FABP3 and / or IL1RL1 in the sample;

[0026] Processing unit: used to obtain results of predicting the risk of irAEs / predicting the severity of irAEs based on the expression levels of FABP3 and / or IL1RL1.

[0027] Advantages and beneficial effects of the present invention:

[0028] This application found through the discovery cohort that FABP3 can well predict the occurrence and severity of irAEs, and further verified the predictive performance of FABP3 through the validation cohort, and constructed a risk prediction model for predicting irAEs. The research of this application is expected to improve the management of irAEs and provide cancer patients with more personalized, effective and safe immunotherapy options. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 is a graph of proteomics data quality and clinical relevance, where 1a is a flowchart of plasma protein analysis, 1b is a graph of the number of proteins identified during the quality control process, 1c is a graph of the number of comparable proteins in different irAEs groups, 1d is a graph of the functional relationship between the mean value of protein intensity and the abundance level in the plasma proteome, 1e is a graph of the coefficient of variation (CV) of each protein in different irAEs groups, 1f and 1g are graphs of scores of dimensions 1 and 2 based on PCA and PLS-DA of comparable proteins, 1h is a graph of the correlation between clinical information and proteomic analysis of irAEs severity assessed by PCA / PLS-DA scores, 1i and 1j are graphs of secretion and tissue-specific proteins identified in samples based on HPA annotations;

[0030] Figure 2 2a and 2b are Venn diagrams of proteins with significant trend differences with irAEs severity determined by two-sided Jonckheere-Terpstra trend test and Kruskal-Wallis test, 2c is a clustering heat map of proteins with trend differences between different irAEs groups, 2d is a fold change scatter plot of proteins with strong or weak trend differences, 2e is a fold change scatter plot of proteins with significant differences between any two irAEs groups or between the control group and the mild group, 2f is a bar chart of KEGG and Reactome terms that are significantly reduced (blue) or increased (red) in trend proteins associated with irAEs severity (FDR P < 0.05 and gene ratio > 0.1) compared with various subgroups;

[0031] Figure 33a and 3b are the relationship diagrams of irAEs network proteins detected in consensus protein co-expression network analysis, 3c is the distribution diagram of proteins detected in irAEs network modules with trend differences, 3d is the clustering heat map of important proteins in co-expression module (c), 3e is the significant correlation diagram of KEGG and Reactome terms with proteins with strong trends in irAEs severity detected in M1 and M4 (terms are marked as decreased (blue) or increased (red), FDR P < 0.05 and gene ratio > 0.1);

[0032] Figure 4 4a and 4c are the comparison diagrams of the model performance of ProIRAE and the single protein model, and 4b and 4d are the distribution diagrams of the predicted risk scores of the selected irAEs severity monitoring models in the discovery cohort and the validation cohort;

[0033] Figure 5 It is a correlation analysis diagram between proteomic profiles and clinical characteristics, among which 5a is a similarity diagram between patient protein profile samples, 5b is a correlation difference distribution diagram between protein profiles and different clinical characteristics under different irAEs groups based on PCA score evaluation, 5c and 5d are correlation diagrams between secretory proteomics and organ-specific proteomics and irAEs clinical characteristics evaluated based on PCA / PLS-DA scores;

[0034] Figure 6 Bar chart of GO enrichment analysis terms that were significantly decreased (blue) or increased (red) in trend proteins (FDR P < 0.05 and gene ratio > 0.1) associated with the severity of irAEs compared with various subgroup comparisons;

[0035] Figure 7 is a consensus protein co-expression network analysis, wherein 7a is a determination diagram of the soft threshold power (β) in the consensus protein co-expression network analysis, 7b is a protein dendrogram based on dissimilarity measurement, 7c is an analysis diagram of the impact of missing values ​​(threshold 10-40% missing) on ​​the evaluation of irAEs network modules (dashed blue (1.96) and red (10) indicate weak statistical significance and high statistical significance, respectively), 7d is a box plot of the differences between the original module characteristic proteins (the first principal component of module protein expression) in the irAEs-related modules between different irAEs groups, 7e is a correlation analysis diagram between the original module characteristic proteins and the module characteristic proteins generated by the proteins selected from the hypergeometric test (3c), and 7f is a correlation analysis diagram between potential biomarkers and consensus network module characteristic proteins;

[0036] Figure 8 This is a bar chart of GO enrichment analysis terms that are significantly reduced (blue) or increased (red) in proteins enriched in the irAEs-related module (FDR P < 0.05 and gene ratio > 0.1);

[0037] Fig. 9 It is a graph of related proteins in KEGG and Reactome entries that are significantly enriched in M1 and M4 modules with high stability;

[0038] Fig.10 This is a model hyperparameter diagram based on five-fold cross-validation of the discovery set data;

[0039] Fig.11 It is an evaluation diagram of the efficacy of using FABP3 protein to predict the severity of irAEs, among which, 11a is the correlation diagram and corresponding scatter plot between the results of FABP3 protein abundance quantification using mass spectrometry (MS), and 11b is the evaluation diagram of the efficacy of ProIRAE model in monitoring the severity of irAEs in the discovery cohort and validation cohort. DETAILED DESCRIPTION

[0040] The definitions of some terms used in this specification are provided below. Unless otherwise defined, all technical and scientific terms used herein generally have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0041] The present invention provides the use of a reagent for detecting the expression level of FABP3 in the preparation of a product for predicting the risk of irAEs / predicting the severity of irAEs.

[0042] The reagent is selected from any of the following groups:

[0043] (1) an oligonucleotide probe that specifically recognizes the FABP3 gene;

[0044] (2) Primers that specifically amplify the FABP3 gene;

[0045] (3) A binding agent that specifically binds to the protein encoded by the FABP3 gene.

[0046] In some embodiments, FABP3 includes wild type, mutant or fragment thereof. The term encompasses full-length, unprocessed FABP3, as well as any form of FABP3 derived from processing in cells, and variants of FABP3 (e.g., splice variants or allelic variants). The term encompasses, for example, human FABP3 and FABP3 from any other vertebrate source, including mammals, such as FABP3 of primates and rodents (e.g., mice and rats), Gene ID: 2170.

[0047] In some embodiments, primer refers to a short nucleic acid molecule, such as a DNA oligonucleotide, such as a sequence of at least 15 nucleotides, which can anneal with a complementary target nucleic acid molecule by nucleic acid hybridization to form a hybrid between the primer and the target nucleic acid chain. The primer can be extended along the target nucleic acid molecule by a polymerase. Therefore, the primer can be used to amplify the target nucleic acid molecule, wherein the sequence of the primer is specific for the target nucleic acid molecule, and the primer hybridizes with the target nucleic acid molecule under very high stringency hybridization conditions.

[0048] Can be by PCR, real-time PCR or other nucleic acid amplification methods known in the art, use primer pair to amplify nucleic acid sequence.Upstream or forward primer is the primer in the 5' direction relative to the reference point on the nucleic acid sequence.Downstream or reverse primer is the primer in the 3' direction relative to the reference point on the nucleic acid sequence.In general, comprise at least one forward and one reverse primer in the amplification reaction.

[0049] In some embodiments, a probe refers to a molecule that can bind to a specific sequence or subsequence or other portion of another molecule. Unless otherwise indicated, a probe generally refers to a polynucleotide probe that can bind to another polynucleotide (often referred to as a target polynucleotide) through complementary base pairing. Depending on the stringency of the hybridization conditions, the probe can bind to a target polynucleotide that lacks complete sequence complementarity with the probe. The probe can be labeled directly or indirectly. Hybridization methods include, but are not limited to, solution phase, solid phase, mixed phase, or in situ hybridization assays.

[0050] The reagent also includes a detectable label.

[0051] In some embodiments, detectable label refers to any composition detectable by fluorescent, spectroscopic, photochemical, biochemical, immunological, electrical, optical or chemical means, including but not limited to UV-Vis labels, near infrared labels, fluorescent groups, phosphorescent groups, magnetic spin resonance labels, photosensitizers, photocleavable moieties, chelating centers, heavy atoms, radioisotopes, isotope detectable spin resonance labels, paramagnetic moieties, chromophores, luminophores.

[0052] In some embodiments, the fluorescent group includes but is not limited to FAM (Carboxyfluorescein, green fluorescence), FITC (fluorescein isothiocyanate), TET (tetrachloro-6-carboxyfluorescein), HEX (hexachloro-6-methylfluorescein), JOE (2,7-dimethyl-4,5-dichloro-6-carboxyfluorescein), rhodamine (Rhodamine dyes such as R110, TAMRA, Texas Red, etc.), ROX, AlexaFluor dyes (such as Alexa 350, Alexa 405, Alexa 430, Alexa 488, Alexa500, Alexa 514, Alexa 532, Alexa 546, Alexa 555, Alexa 568, Alexa594, Alexa 610, Alexa 633, Alexa 635, Alexa 647, Alexa 660, Alexa 680, Alexa700, Alexa 750, Alexa 790), ATTO dyes (such as ATTO 390, ATTO 425, ATTO 465, ATTO 488, ATTO 495, ATTO 514, ATTO 520, ATTO 532, ATTO Rho6G, ATTO 542, ATTO 550, ATTO 565, ATTO Rho3B, ATTO Rho11, ATTO Rho12, ATTO Thio12, ATTO Rho101, ATTO 590, ATTO 594, ATTO Rho13, ATTO 610, ATTO 620, ATTO Rho14, ATTO 633, ATTO 647, ATTO 647N, ATTO 655, ATTO Oxa12, ATTO 665, ATTO 680, ATTO 700, ATTO 725, ATTO 740), DyLight dyes, cyanine dyes (such as Cy2, Cy3, Cy3.5, Cy3b, Cy5, Cy5.5, Cy7, Cy7.5), FluoProbes dyes, SulfoCy dyes, Seta dyes, IRIS dyes, SeTau dyes, SRfluor dyes, Square dyes.

[0053] The present invention provides a system for predicting the risk of irAEs / predicting the severity of irAEs, the system comprising:

[0054] Acquisition unit: used for acquiring the expression level of FABP3 and / or IL1RL1 in the sample;

[0055] Processing unit: used to obtain results of predicting the risk of irAEs / predicting the severity of irAEs based on the expression levels of FABP3 and / or IL1RL1.

[0056] In some embodiments, the implementation of the system may include performing or completing the selected tasks manually, automatically, or in combination thereof. Moreover, actual instruments and devices according to embodiments of the system of the present invention may implement multiple selected tasks using an operating system by hardware, by software, or by firmware, or by a combination thereof.

[0057] For example, the hardware for performing the selected task can be a chip or a circuit. As software, the selected task can be implemented as a plurality of software instructions executed by a computer using any suitable operating system. In the present invention, one or more tasks according to the exemplary embodiments of the method and / or system as described in the present invention can be performed by a processing unit, such as a computing platform for executing a plurality of instructions. Optionally, the processing unit includes a volatile memory for storing instructions and / or data, and / or a non-volatile memory for storing instructions and / or data, for example, a magnetic hard disk and / or a removable medium. Optionally, a network connection is also provided. A display and / or a user input device such as a keyboard or a mouse are also optionally provided.

[0058] The invention will be further described below in conjunction with specific examples. It should be understood that the specific embodiments described herein are presented by way of example and are not intended to limit the invention. The main features of the invention may be used in a variety of embodiments without departing from the scope of the invention.

[0059] Example

[0060] 1. Experimental methods

[0061] Patient cohort and sample collection

[0062] This study is a prospective discovery-validation cohort study conducted at the Affiliated Cancer Hospital of Harbin Medical University from August 2020 to April 2024. This study has been approved by the Ethics Committee of the Affiliated Cancer Hospital of Harbin Medical University (approval number: kyk2020003) and follows the guidelines of the Declaration of Helsinki. Informed consent was obtained from all participants before enrollment.

[0063] A total of 65 patients who received anti-PD-1 / anti-PD-L1 blockade therapy were recruited and divided into two cohorts: a discovery cohort and a validation cohort. The discovery cohort included 47 patients recruited between August 2020 and February 2023, while the validation cohort included 18 patients recruited between April 2023 and April 2024.

[0064] Eligible patients were aged between 30 and 75 years and had received anti-PD-1 or anti-PD-L1 therapy (monotherapy or combination) for at least one week. Patients with a history of immune system diseases (including AIDS, systemic lupus erythematosus, hyperthyroidism, or Cushing's syndrome) or who had undergone an organ transplant in the past six months were not included in the study. All adverse reactions were evaluated according to the National Comprehensive Cancer Network (NCCN) guidelines version 1.2020. These adverse events affected different organ systems, including but not limited to the skin, liver, lungs, kidneys, nervous system, and cardiovascular system. The demographic and clinical data of participants collected included sex, age, body mass index (BMI), treatment duration, cancer type, and stage. According to the severity of irAEs, patients were divided into three groups: 1) control group: patients who did not experience any irAEs; 2) mild group: patients with irAEs graded less than grade 3; 3) severe group: patients with irAEs graded 3 or above. Plasma samples were obtained from all participants. In the mild and severe groups, plasma samples were collected on the first day of irAE diagnosis.

[0065] Plasma protein extraction and trypsin digestion

[0066] Use EDTA tube (BD Beckton Dickinson, Franklin Lakes, NJ, US) and centrifuged at 1000 × g for 10 min at 4°C within 2 h of collection. Plasma was separated at -80°C using a standardized protocol and then centrifuged at 12,000 × g for 10 min at 4°C to remove cellular debris. The supernatant was transferred to a new centrifuge tube. The supernatant was centrifuged using Pierce TM The top 14 abundant proteins were removed using the Top 14 Abundant Protein Depletion Spin Columns Kit (ThermoFisher Scientific). Protein concentration was determined using the BCA assay kit according to the manufacturer's instructions.

[0067] An equal amount of protein was used for digestion in each sample. Sample amounts were adjusted with lysis buffer to ensure consistency. The protein solution was reduced with 5 mM dithiothreitol at 56 °C for 30 min and then alkylated with 11 mM iodoacetamide at room temperature in the dark for 15 min. The alkylated sample was transferred to an ultrafiltration device and centrifuged at 12,000 × g for 20 min at room temperature. The sample was washed three times with 8 M urea and then three times with buffer solution. Trypsin was added at a 1:50 enzyme-protein ratio and digested overnight. The resulting peptides were collected by centrifugation at 12,000 × g for 10 min at room temperature and then washed once more with double distilled water. The combined peptide solution was used for liquid chromatography-tandem mass spectrometry (LC-MS / MS) analysis.

[0068] LC-MS / MS analysis

[0069] The peptides were dissolved in solvent A (0.1% formic acid and 2% acetonitrile in water) and then separated using a Vanquish ultra-high performance liquid chromatography system. The peptides were separated using a gradient method, with the concentration of solvent B (0.1% formic acid and 90% acetonitrile) increasing from 7% to 20% over 16 minutes; 20% to 32% over 8 minutes; then to 80% over 3 minutes and maintained at 80% for the last 3 minutes. The flow rate was maintained at 500 nL / min. After separation, the peptides were ionized using a nanospray ionization (NSI) source and analyzed using an Orbitrap Exploris 480 mass spectrometer. The ion source voltage was set to 2.3 kV, and the FAIMS compensation voltage was set to -45 V and -70 V. High-resolution detection and analysis of precursor and fragment ions were performed using the Orbitrap. The full MS scan resolution was set to 30,000 with a scan range of 390–810 m / z; the MS / MS scan resolution was set to 30,000 with a fixed start point of 200 m / z.

[0070] Data acquisition was performed in data independent acquisition (DIA) mode. After each full mass spectrometry scan, multiple consecutive m / z peptide ion windows were selected and fragmented in the high energy collision dissociation (HCD) collision cell. Fragmentation was performed at normalized collision energies of 25%, 30%, and 35%, followed by MS / MS analysis. To optimize the efficiency of mass spectrometry analysis, the automatic gain control (AGC) target was set to 3E6 and the maximum injection time was set to automatic.

[0071] Peptide identification and protein quantification

[0072] The resulting DIA data were searched against the UniProt human protein database (updated 2023-01-03, 20,389 entries) using Spectronaut (v17) and the Pulsar search engine (default settings). A reverse decoy database was included in the search to calculate the false discovery rate (FDR) and to account for random hits. Trypsin / P was specified as the cleavage enzyme, and a maximum of two missed cleavages were allowed. Cysteine ​​aminomethylation was specified as a fixed modification, while acetylation at the protein N-terminus and oxidation of methionine residues were specified as variable modifications. The FDR threshold for peptide spectrum matches (PSMs), peptide and protein identification and quantification was strictly set to 1%.

[0073] Proteomics data quality control and preprocessing

[0074] Protein abundance values ​​less than 10 were marked as missing values ​​(NA). Proteins identified in less than 50% of the samples were not further analyzed, and the remaining missing value proteins were estimated using the R software package "DreamAI" with a parameter setting of k = 10 and the k-nearest neighbor (k-NN) method. After estimation, protein intensities were log-2 transformed for subsequent statistical and bioinformatics analysis.

[0075] Differential protein analysis

[0076] Differential analysis of protein expression was performed to identify proteins with significant trends between the control group, mild and severe irAEs groups. Proteins that were consistently up- or down-regulated were identified using the Kruskal-Wallis test and the Jonckheere-Terpstra test to assess monotonic trends in protein expression. The direction of these trends was determined by the median protein intensity of each group. The Wilcoxon rank sum test was used to compare protein expression between the control group and the mild group, the mild group and the severe group, the control group and the severe group, and the protein expression with strong trend changes between irAEs and no irAEs, and severe irAEs and non-severe irAEs. If the FDR < 0.05, the protein was considered to have a significant abundance difference under a given condition.

[0077] Functional annotation and enrichment analysis

[0078] Secretory proteins and tissue-specific proteins were annotated according to the Human Protein Atlas (HPA). Gene Ontology (GO), KEGG pathway, and Reactome enrichment analysis of differentially expressed proteins were performed using the R packages “clusterProfiler” and “ReactomePA” with an FDR < 0.05.

[0079] Weighted Correlation Network Analysis (WGCNA)

[0080] Protein co-expression modules were identified by unsupervised clustering using the blockwiseModules subroutine in the R package “WGCNA” with the following settings: soft power = 6, correlations were calculated using bidweight mid-correlations (bicor), deepSplit = 4, minimum module size was 14, merged cut height was 0.07, average TOM denominator, signed network, minimum kME of 0.4 for retaining proteins in a module, and redistribution threshold was P < 0.05.

[0081] To reduce the number of modules with highly similar module signature proteins (representative abundance values ​​of each module), the WGCNA::mergeCloseModules() function was used. Proteins were reassigned based on Spearman correlation with a maximum similarity threshold of 0.25. Module preservation at different levels of proteome loss was assessed using the WGCNA::modulePreservation() function, and unassigned proteins were marked as M0 and annotated in gray.

[0082] Enzyme-linked immunosorbent assay (ELISA)

[0083] Plasma protein levels were quantified using commercially available ELISA kits according to the manufacturer's instructions. The specific kits used in this study were as follows: IL1RL1 (CSB-E13789h), ADM (CSB-E09146h), FABP3 (CSB-E09185h), TNC (CSB-E13125h), TNFIP6 (CSB-EL023959HU), VSIG4 (CSB-EL025928HU), PLIN3 (MM-63554H2), and VWCE (MM-51800H2). All experimental procedures were performed according to the protocols provided by the kit manufacturers.

[0084] Machine learning-based protein model for irAEs risk prediction

[0085] Feature importance analysis was performed using the variable importance in projection (VIP) method based on protein abundance data derived from mass spectrometry (MS) and enzyme-linked immunosorbent assay (ELISA) to estimate the contribution of each variable in predicting the risk of irAEs. To build a predictive model, an ordinal logistic regression model was trained in the discovery cohort using the best combination of variables. Hyperparameters were optimized using grid search and five-fold cross validation.

[0086] Statistical analysis

[0087] All statistical analyses were performed using R (v4.2.0) and related software packages. Intergroup comparisons of continuous variables were performed using the Wilcoxon rank sum test or the Kruskal Wallis test, as appropriate. Categorical variables were analyzed using the Fisher exact test. For model evaluation, the area under the receiver operating characteristic curve (AUROC) and the area under the precision-recall curve (AUPR) were calculated to evaluate the performance of the machine learning model.

[0088] Experimental Results

[0089] Study design and clinical characteristics of the irAEs cohort

[0090] To investigate plasma proteomic patterns and identify potential non-invasive biomarkers of immune checkpoint blockade-related irAEs, comprehensive proteomic analysis of plasma samples was performed. A total of 65 patients treated with anti-PD-1 / PD-L1 therapy were included in this study and divided into three groups according to the severity of irAEs: control group (no irAEs), mild group (irAEs grade <3), and severe group (irAEs grade ≥3). Figure 1 a), patients were further divided into a discovery cohort consisting of 47 patients (11 controls, 14 mild groups, and 22 severe groups) and a validation cohort consisting of 18 patients (7 controls, 5 mild groups, and 6 severe groups). The discovery cohort was used for proteomic analysis, biomarker identification, and model construction, while the validation cohort was used to evaluate the generalizability of the model.

[0091] The average age of the patients ranged from 56.6 to 65.4 years, and the body mass index ranged from 20.9 to 25.7. The study adopted a pan-cancer approach because all participants had irAEs due to disruption of the immune signaling pathways targeted by PD-1 / PD-L1 therapy. The most common cancer types among participants were lung cancer and liver cancer, as well as 13 other less common cancers.

[0092] Plasma proteome analysis of irAEs

[0093] Using the Deep Blood+DIA technology, a total of 15,745 peptides and 2,304 proteins were identified in 47 plasma samples of the discovery cohort ( Figure 1 b). After quality control, 1,459 proteins were retained for further analysis. Comparative proteomic analysis showed that there was no significant difference in proteome coverage between the control group (median, 1,316 proteins) and the mild irAEs group (median, 1,311 proteins) (Wilcoxon rank sum test, P = 0.956) ( Figure 1c). However, there was a significant difference between the severe irAEs group (median, 1,339 proteins) and the other two groups (Wilcoxon rank-sum test, P = 0.027). Protein intensities spanned six orders of magnitude, with the ten most abundant proteins accounting for 39% to 45% of the total abundance of plasma proteins ( Figure 1 d). Notably, the variability of protein expression increased with increasing severity of irAEs, with the median coefficient of variation increasing from 6.5% in the control group to 9.9% in the severe irAEs group, indicating increased proteomic heterogeneity ( Figure 1 e and Figure 5 a).

[0094] Unsupervised principal component analysis (PCA) and supervised partial least squares discriminant analysis (PLS-DA) were used to compare and observe the plasma proteome profiles, and the results showed that there were significant differences among the three irAEs groups ( Figure 1 fh and Figure 5 b). Further annotation of the HPA identified 657 secreted proteins (45.0% of the analyzed proteins), of which 300 (45.7%) were actively secreted into the blood and 141 (21.5%) were transmembrane proteins ( Figure 1 i and Figure 5 c). In addition, 269 tissue-specific proteins were found (18.4% of the analyzed proteins), most of which were derived from the liver (124 proteins, 46.1%), implying a significant contribution of the liver to the plasma proteome ( Figure 1 j and Figure 5 d). In different irAEs groups, both secreted proteins and tissue-specific proteins showed consistent patterns and had no significant correlation with other clinical characteristics. In conclusion, these findings suggest that plasma proteomics can effectively reflect the onset and severity of irAEs.

[0095] Identification of potential plasma protein biomarkers for irAEs

[0096] To further investigate the changes in plasma proteome associated with the onset and severity of irAEs, a differential trend analysis was performed, and it was found that the expression of 217 proteins changed significantly in different irAEs severity. Among them, the expression of 184 proteins increased significantly, and the expression of 33 proteins decreased significantly. In addition, the abundance of 106 proteins changed weakly, of which the abundance of 85 proteins increased with the severity of irAEs, and the abundance of 21 proteins decreased with the severity of irAEs ( Figure 2a). Proteins with a clear strong trend showed significant differences in distinguishing irAEs samples (mild group + severe group) from severe irAEs samples, and the changes between the mild irAEs group and the severe irAEs group were more obvious than those between the control group and the mild irAEs group, indicating that the changes in the proteome became more obvious with the escalation of irAEs severity ( Figure 2 b).

[0097] Enrichment analysis of these dysregulated proteins showed that, although the clustering of samples was relatively consistent, proteins with stronger trends were more strongly regulated in each irAEs group compared with proteins with weaker trends ( Figure 2 c, 2f, Figure 6 ). Functional annotation using Reactome and KEGG showed that proteins with strong trends were mainly involved in inflammation and chemotaxis (such as neutrophil degranulation, neutrophil interstitium, neutrophil apoptosis, neutrophil degranulation, interleukin signaling, and interleukin-1 family signaling), metabolism and cellular responses (such as amino acid and carbohydrate metabolism, extracellular matrix organization, and cellular response to chemical stress), and immunity and antigen processing (such as class I MHC-mediated antigen processing and presentation and antigen processing: ubiquitination and proteasomal degradation). In contrast, pathways related to complement and coagulation functions and humoral immune functions (such as antigen binding and immunoglobulin secretion) were continuously downregulated, indicating that with the increase in the severity of irAEs, humoral immunity (antibody-mediated) has shifted to cellular immunity (T cell-mediated) ( Figure 2 f and Figure 6 ).

[0098] Several previously identified irAEs biomarkers were also differentially expressed in plasma, including C-reactive protein (CRP, associated with systemic inflammatory response; Kruskal-Wallis FDR P = 0.028), cluster of differentiation 177 (CD177, involved in immune-mediated intestinal diseases; Kruskal-Wallis FDR P = 0.009), and lymphocyte cytoplasmic protein 1 (LCP1; Kruskal-Wallis FDR P = 0.009), although their expression was stable in the control and mild groups ( Figure 2 d). By comparing different irAEs groups, it was found that 8 proteins continued to increase with the increase in the severity of irAEs ( Figure 2 b, 2e). These proteins highlight their potential as non-invasive biomarkers for predicting and monitoring the risk and progression of irAEs in patients receiving anti-PD-1 / PD-L1 therapy.

[0099] Clinically relevant functional protein modules

[0100] To identify co-regulated proteins and their association with irAEs, WGCNA analysis of the proteomic data of plasma samples revealed 17 different protein modules, ranging in size from 255 proteins in M0 to 15 proteins in M16 ( Figure 7 a, 7b). Robustness analysis shows that these modules remain highly stable under different data integrity thresholds (10% to 50%) ( Figure 7 c). In addition, five modules were significantly associated with the risk and severity of irAEs, indicating that they were involved in the progression of irAEs regardless of other clinical characteristics ( Figure 3 a, 3b and Figure 7 d). Unsupervised hierarchical clustering of 704 proteins in four key protein modules (M1, M2, M4, and M16) revealed that the irAEs in the control, mild, and severe groups were significantly different ( Figure 3 d), indicating that these modules are closely related to the clinical manifestations of irAEs.

[0101] Then, 217 proteins with strong differential trends (identified in the previous analysis) were mapped to these functional protein modules using the hypergeometric test. The results showed that three modules (M1, M2, and M16) were significantly enriched for proteins with an up-regulated trend, with M1 enriched 2.34 times, M2 enriched 4.27 times, and M16 enriched 3.96 times. In contrast, module M4 was enriched for down-regulated proteins by as much as 13.96 times ( Figure 3 c). These findings highlight the close association between module composition and irAEs progression, as the recalculated signature proteins showed high correlations with the original module signature proteins, ranging from 0.93 to 0.99 ( Figure 7 e). Functional enrichment analysis was performed based on the KEGG and Reactome databases to identify the main biological processes of each protein module. It was found that M1 was significantly involved in inflammatory and chemotaxis pathways and matrix organization, indicating increased T cell activity; M2 was mainly related to antigen-antibody interactions and disease-related pathways; while M4 showed a downward trend in complement and coagulation-related functions ( Figure 3 d, Figure 8 , Fig. 9 ).

[0102] Development and validation of a plasma protein-based prediction model for irAEs

[0103] To evaluate the clinical applicability and reproducibility of the eight potential protein biomarkers identified, plasma protein levels were measured in previous patients using enzyme-linked immunosorbent assay. The plasma protein levels of most candidate biomarkers measured by ELISA correlated well with those measured by MS. Then, feature selection was performed using the VIP method to identify the most informative protein biomarkers to establish a predictive model. The proteomic data (MS and ELISA) of the discovery cohort showed two highly informative biomarkers: interleukin 1 receptor-like 1 (IL1RL1) and fatty acid binding protein 3 (FABP3), which were selected into the plasma protein-based irAEs risk prediction (PROIRAE) model. The PROIRAE model was constructed using ordinal regression and optimized with five-fold cross-validation and grid search for hyperparameter adjustment ( Fig.10 ). The PROIRAE model showed strong predictive performance. For the discovery cohort, PROIRAE achieved an AUROC of 0.932 and an AUPR of 0.980 in identifying patients at risk of developing irAEs, and an AUROC of 0.980 and an AUPR of 0.982 in identifying patients at risk of developing severe irAEs ( Figure 4 a). The risk score predicted by the PROIRAE model showed a clear stratification of the severity of irAEs ( Figure 4 b).

[0104] To verify the universality of PROIRAE, it was applied to an independent cohort of 18 patients. The PROIRAE model maintained strong predictive performance, with an AUROC of 0.727 and an AUPR of 0.880 in distinguishing non-irAEs from irAEs patients; and an AUROC of 0.972 and an AUPR of 0.979 in predicting severe irAEs ( Figure 4 c). The distribution of predicted risk scores in the control group, mild group, and severe group in the validation cohort also showed obvious stratification, further validating the ability of the model to reliably distinguish patients with different degrees of irAEs ( Figure 4 d). In addition, the combination of IL1RL1 and FABP3 in the PROIRAE model outperformed the predictive ability of either protein biomarker alone, highlighting the synergistic role of these two protein biomarkers in predicting the risk and severity of irAEs.

[0105] The description of the above embodiments is only used to understand the method and core idea of ​​the present invention. It should be pointed out that, for those skilled in the art, several improvements and modifications can be made to the present invention without departing from the principle of the present invention, and these improvements and modifications will also fall within the scope of protection of the claims of the present invention.

Claims

1. A product for predicting the risk of irAEs / predicting the severity of irAEs, characterized in that: The product includes a reagent for detecting the expression level of FABP3.

2. The product according to claim 1, characterized in that The product includes a chip, a test kit or a nucleic acid membrane strip.

3. The product according to claim 2, characterized in that The kit also includes a container; Preferably, the kit further comprises instructions; Preferably, the kit further comprises a buffer; Preferably, the kit further comprises packaging materials.

4. A model for predicting the risk of irAEs / predicting the severity of irAEs, characterized in that: The model includes FABP3; Preferably, the model further comprises IL1RL1.

5. Application of reagents for detecting FABP3 expression levels in the preparation of products for predicting the risk / predicting the severity of irAEs.

6. The use according to claim 5, characterized in that: The reagent is selected from any of the following groups: (1) an oligonucleotide probe that specifically recognizes the FABP3 gene; (2) Primers that specifically amplify the FABP3 gene; (3) a binding agent that specifically binds to the protein encoded by the FABP3 gene; Preferably, the product further comprises a reagent for detecting the expression level of IL1RL1.

7. The use according to claim 6, characterized in that: The reagent also includes a detectable label; Preferably, the detectable marker comprises a UV-Vis label, a near-infrared label, a fluorescent group, a phosphorescent group, a magnetic spin resonance label, a photosensitizer, a photocleavable part, a chelating center, a heavy atom, a radioisotope, an isotope detectable spin resonance label, a paramagnetic part, a chromophore, a luminophore.

8. Application of FABP3 in constructing a model for predicting the risk / severity of irAEs.

9. Application of FABP3 in building a system for predicting the risk / severity of irAEs.

10. A system for predicting the risk of irAEs / predicting the severity of irAEs, characterized in that: The system comprises: Acquisition unit: used for acquiring the expression level of FABP3 and / or IL1RL1 in the sample; Processing unit: used to obtain results of predicting the risk of irAEs / predicting the severity of irAEs based on the expression levels of FABP3 and / or IL1RL1.

Images