Dissecting Non-Small Cell Lung Cancer (NSCLC) with Blood Proteomics - From Surgical to Immunotherapeutic Responses

Dissecting Non-Small Cell Lung Cancer (NSCLC) with Blood Proteomics - From Surgical to Immunotherapeutic Responses | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Dissecting Non-Small Cell Lung Cancer (NSCLC) with Blood Proteomics - From Surgical to Immunotherapeutic Responses Vahid Yaghoubi Naei, Aaron Kilgallon, Gwendoline Mendes, Sanjay Dutta, and 10 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7118825/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 9 You are reading this latest preprint version Abstract Objectives. Non-small cell lung cancer (NSCLC) remains a leading cause of cancer-related mortality, with limited biomarkers to guide surgical and immunotherapeutic intervention. This study aimed to identify plasma proteomic signatures associated with surgical resection, recurrence, immunotherapy response, and survival by leveraging both high-plex and high-sensitivity proteomics technologies. Methods. Two complementary plasma proteomics platforms, SomaScan (quantifying 7596 proteins) and NULISAseq (quantifying 250 inflammation-related proteins), were used to profile the blood at 87 timepoints from 56 NSCLC patients. Samples were collected longitudinally: pre- and post-surgery (n = 20), pre- and post-immune checkpoint inhibitor (ICI) therapy (n = 11), and at baseline prior to ICI (n = 25). Proteomics data were analysed using adaptive Lasso regression (ALasso) and the Stabl algorithm for robust selection of a minimal number of differentially expressed proteins that collectively modelled the clinical classification or response. Results. We identified 21 differentially detectable plasma proteins across surgery and ICI treatment. Surgical resection induced measurable changes in plasma proteins, notably higher circulating MUC16 and lower circulating IL36G post-surgery in non-recurrent patients. Recurrent patients had higher plasma levels of COX7A2L, FGF19 and SPOCK2 post-surgery, and lower FCER2, FCRLA and SLITRK2. Patients who responded to ICI therapy had lower levels of baseline IL-6, CCL19, IL-2RA, CD200R1, CRP, LIF, PDCD1, CCL7 and SPP1 prior to ICI therapy, highlighting associations between systemic inflammation and immune regulation. CEACAM5, PTX3, FGF23, and AREG were elevated in patients with worse clinical outcomes and poorer overall survival. Cross-platform comparisons underscored the complementary strengths of SomaScan’s broad coverage and NULISA’s sensitivity in detecting low-abundance, clinically relevant proteins. Conclusions. This integrated plasma proteomics study reveals distinct protein signatures associated with NSCLC treatment response and prognosis. These findings support the utility of non-invasive, blood-based proteomic assays for defining biomarker discovery in NSCLC. Health sciences/Biomarkers Biological sciences/Cancer Biological sciences/Immunology Health sciences/Oncology Plasma proteomics NSCLC NULISA SomaScan immunotherapy biomarkers Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Lung cancer, in particular non-small cell lung cancer (NSCLC), remains among the leading causes of cancer-related deaths worldwide and is the second most frequently diagnosed cancer [ 1 , 2 ]. Detecting lung cancer at an early stage is a critical need to improve therapeutic options and long-term health outcomes [ 3 ]. However, NSCLC, the most common form of lung cancer, is often identified at advanced stages, where patient prognosis is often poor. Early detection is thus critical to improving survival. Currently, low-dose computed tomography (LDCT) screening of high-risk individuals (like heavy smokers) can identify early-stage lung cancers and has been shown to reduce lung cancer mortality [ 4 ]. However, LDCT’s impact is limited by significant drawbacks: screening yields a high rate of false positives (especially in older ages) leading to higher costs and patient anxiety [ 5 ]. Additionally, overdiagnosis of indolent lesions and cumulative radiation exposure pose ongoing concerns [ 6 ]. These limitations, combined with the fact that many at-risk individuals are ineligible or do not undergo screening, indicate the importance of better non-invasive diagnostic tools. Plasma proteomics offers a non-invasive means to identify these changes and assist early diagnosis, prognostication, and response to therapy monitoring [ 7 ]. The plasma proteome encompasses a wide range of proteins with different concentrations, reflecting the dynamic state of biomarkers and can be a good proxy for determining tumour dynamics [ 8 ]. A recent study identified a panel of three plasma proteins that can predict NSCLC risk up to 10 years before diagnosis (achieving ≈ 0.90 area under the curve (AUC) in 5-year predictions) [ 9 ]. Likewise, longitudinal proteomic analyses in patients receiving therapy have linked changes in specific plasma proteins to treatment responses. rising levels of soluble PD-1 in the bloodstream during immune checkpoint inhibitor therapy have been shown to correlate with tumour PD-L1 status and improved patient outcomes [ 10 ]. These advances highlight the advantages of plasma proteomics as a complementary strategy to imaging, with the potential to improve early NSCLC diagnosis and to enable proactive treatment response monitoring for better clinical management [ 11 ]. Precise and unbiased affinity-based methods for large-scale proteome analysis like SomaScan (SomaLogic, US) and Nucleic Acid-Linked Immuno-Sandwich Assay (NULISA) (Alamar Biosciences, Canada) have been recently developed, demonstrating high sensitivity and specificity for a wide array of proteins across the entire dynamic range in complex fluids, like serum or plasma. These advancements address the limitations of traditional and even complex mass spectrometry technologies in detecting novel disease biomarkers [ 12 , 13 ]. In this study, we aim to investigate the plasma proteomes of NSCLC patients using SomaScan v4.1, which comprises reagents to detect 7596 plasma proteins including cardiovascular, inflammation and immune response, metabolic diseases, oncology, neurology, cytokines, and respiratory disease biomarkers, and Alamar Biosciences NULISAseq Inflammation Panel, which detects 250 biomarkers, before and after surgery, and before and after immune checkpoint immunotherapy, to discover potential prognostic and diagnostic biomarkers. We utilized the feature selection method Stabl in combination with linear modeling to identify protein signatures associated with plasma proteome changes induced by adjuvant surgery, proteins associated with disease recurrence, proteins prognosticative of response to immunotherapy, and proteins associated with patient survival [ 14 ]. Methods Patient cohort Ethics approval for this study was obtained from the Metro South Health District Human Research Ethics Committee under the National Health and Medical Research Council guidelines (HREC/11/QPAH/331) to collect samples from the Princess Alexandra Hospital. Patients who were selected for this study were treatment-naive and without prior cancer diagnosis within the last 5 years. This study has been ratified by the Queensland University of Technology Human Research Ethics Committee. Patients were grouped based on treatment modality (Fig. 1 ). For 20 patients who underwent tumour resection surgery, peripheral blood samples were collected within one week prior to surgery and again 2–4 weeks post-surgery. All patients in this group were enrolled with signed informed consent and followed prospectively for clinical outcomes over a period of three to five years. For 36 patients who received immunotherapy (Pembrolizumab, Atezolizumab and Durvalumab), or combination therapy including chemotherapy (Carboplatin, Pemetrexed, Cisplatin, Paclitaxel, Gemcitabine and Docetaxel), targeted therapy (Osimertinib and Crizotinib) and immunotherapy, blood samples were collected at baseline (on the day of or the day before the first round of systemic therapy) for 25 patients, and at baseline and 3–6 months after treatment initiation for 11 patients. Patient responses to immunotherapy were assessed based on RECIST 1.1 criteria [ 15 ]. All participants were recruited from the Princess Alexandra Hospital thoracic and lung cancer clinic between March 2019 and August 2024. Blood sample processing Blood samples were collected from 56 patients at 87 timepoints in K2EDTA-coated tubes (BD Diagnostics, US), immediately mixed, and transported to the lab at room temperature. Plasma separation was completed within 4 hours of collection by centrifugation at 2200×g for 15 minutes at 4°C with the break off. The samples were then aliquoted and stored at − 80°C prior to blood proteomic profiling. Blood Proteomic profiling of NSCLC plasma samples SomaScan (Somalogic, US) Matched plasma samples from 20 patients pre- and post-tumour resection (40 samples in total, 55 µL/sample) were protein profiled using the SomaScan v4.1 assay. Plasma was analysed in the SomaLogic Laboratory in Colorado using a biomarker panel with 7596 SOMAmers (7301 have available UniProt IDs). Samples were randomised and run on 96-well plates alongside external control samples (calibrators, quality control (QC), and buffer). Serial dilutions (0.05%, 0.5%, and 20%) were used to optimise the detection of protein targets, with the full panel of SOMAmers present in all dilutions, a design consistently applied across all samples. Raw data were standardised using external control samples to adjust for variability in microarrays and discrepancies both within and between plates. This process included adaptive normalisation by maximum likelihood (ANML) to an external reference, aimed at reducing inter-sample variability. The final SomaScan data were provided in both ANML and non-ANML formats in relative fluorescence units (RFU), and log-transformed (log-2) for the main analysis. The limit of detection (LOD) for each SOMAmer was determined using external buffer samples. Quality control was performed by comparing the median of QC samples on each plate to the reference, and a cross-plate QC check measure (pass/flag) was assigned to each SOMAmer. Alamar ARGO HT (Alamar Biosciences, Canada) The Alamar ARGO™ HT system (Model #B31), a fully automated platform designed for multiplex proteomic assays, was used to analyze 25 baseline samples collected before immunotherapy, 11 baseline samples and their matched post-immunotherapy counterparts (22 samples in total), and 17 paired samples collected before and after tumour resection surgery, all of which were matched with the SomaScan assay. Plasma was run on the NULISAseq inflammation Panel 250. Briefly, immunocomplexes were formed using paired oligonucleotide-conjugated antibodies, followed by sequential capture steps with oligo-dT and streptavidin beads, ligation of DNA reporters, and barcoding of samples. The resulting libraries, comprising both target- and sample-specific molecular identifiers, were PCR-amplified, purified, and sequenced using the AVITI platform (Element Biosciences). Each run incorporated multiple layers of quality control, including internal controls (exogenous spike-ins) to normalise individual sample counts, pooled plasma controls to assess intra- and inter-plate consistency, and negative controls (assay buffer only) to evaluate background signal and determine the limits of detection (LOD). Data normalisation followed a multi-step process: raw target counts were first divided by the internal control counts for each sample, then normalised to the median of the inter-plate control for each target and scaled by a factor of 10⁴. A pseudo count of + 1 was added prior to log2 transformation. The final output, expressed as NULISAseq Protein Quantification (NPQ) units, reflects log2-normalised protein expression levels. Plate-specific LODs were determined by calculating the mean plus three standard deviations of the untransformed signal from the negative controls, followed by the same normalisation and log transformation steps. To ensure data quality, any sample with internal control values deviating by more than ± 40% from the plate-wide median was flagged for further evaluation. These samples underwent additional assessment using downstream dimensionality reduction techniques, including principal component analysis and heatmap clustering, to determine their suitability for inclusion in the final dataset. Statistical analysis We used a multitiered statistical framework, including descriptive statistics, univariate tests, and false discovery-corrected feature selection. For feature selection, we used adaptive Lasso (ALasso) regression within the Stabl algorithm, a method established by SurgeCare [ 14 ], and standard stability selection using ALasso [ 16 , 17 ]. Stabl is a statistical framework that uses artificial feature injection, stability-based feature selection, and false discovery control of a False Discovery Proportion (FDP) estimate to find reliable sets of predictive biomarkers from high-dimensional datasets. The final fits to the selected markers were varied on the relevant sample subset and ranged from unweighted Logistic Regression models to Ridge-Regression models and l1-weighted Cox Proportional Hazards models. In each pseudo experiment, data were randomly subsampled, artificial features were created using random permutation, and false discovery rates of uninformative features were estimated using these artificial features. Cross-validation was done using a robust Monte Carlo 5x5 method, with stratification by patient ID to prevent data leakage and to preserve the integrity of comparisons between matched samples from the same patient. The feature selection and the choice of a stability threshold of Stabl were adjusted by evaluating several parameters, such as the final regression regularisation parameters, artificial feature generation approach, and, where false-discovery control failed, signal-to-noise ratio. For each comparison modality, the most stable features, defined as those exceeding the optimized data-driven selection frequency threshold and for which the associated model achieved statistical significance in predictions (P < 0.05), were retained. The area under the receiver operating characteristic curve (AUROC) was used to evaluate the predictive performance of the models. To account for multiple comparisons in the pathway analysis, p-values were adjusted using an FDR correction. K-fold tests in the survival analysis to predict patient response were conducted with four folds and ten repeats. Correlation analysis was performed on matched plasma samples examined via both the SomaScan and NULISA platforms. Shared markers consistent across both assays were chosen, and the Pearson correlation of expression levels were clustered using the distance metric to assess regions of cross-platform panel correlation (Fig. 3 ). Results Surgery-Induced Alterations in the Plasma Proteome Matched patient samples pre- and post-surgery revealed measurable changes in protein expression. The predictive models constructed on the selected proteins demonstrated strong performance, with an AUROC of 0.81 for ALasso (Fig. 2 A) and 0.76 for Stabl (Fig. 2 C). The fold change plot generated with the NULISAseq assay highlighted an opposing trend in the expression of two key markers: Mucin-16 (MUC16) was upregulated after surgery, whereas Interleukin-36 gamma (IL36G) was downregulated (P = 0.01). These markers were consistently identified by both feature selection methods (ALasso and Stabl) as leading predictors distinguishing pre- and post-surgical states (Fig. 2 B, D). From the SomaScan assay using both selection models (Fig. 2 E, G), Thrombospondin-3 (THBS3), Tetranectin (CLEC3B), KxDL motif-containing protein 1 (KDXI), SHC-transforming protein 1 (SHC1), and Transforming growth factor-beta-induced protein ig-h3 (TGFBI) were identified as differentially expressed markers. Differential expression box plots indicated a reduction for CLEC3B, TGFBI, and KDXI, but an increase for THBS3 and SHC1 in post-surgery samples (Fig. 2 F, H). On both platforms, MUC16, IL36G, THBS3, CLEC3B, and TGFBI were chosen by both selection methods. Comparing the NULISA assay with SomaScan protein expression in a univariate analysis revealed that the former had a higher statistical limit for detecting log-fold changes in protein expression, mainly due to its higher sensitivity (Fig. 2 I). Pathway enrichment analysis of the selected proteins highlighted key biological processes affected by surgery, including cytokine activity (P = 0.0093) and signalling (P = 0.031), extracellular compartments (P = 2.46e-06), and cytokine–cytokine receptor interaction (P = 0.0016). Correlating Protein Signatures Across Platforms To evaluate the consistency of protein measurements across technologies, we performed a cross-platform correlation analysis of matched pre- and post-surgery samples. This approach enabled us to identify overlapping protein signals and assess agreement between the SomaScan and NULISA platforms. Cross-platform correlation analysis of the pre- and post-surgery samples run on both platforms identified a cluster of highly correlated proteins that consistently exhibited similarity, although the correlation between both panels generally remained positive. Amongst them, Chemokines (CCL28, CXCL2, CCL17, CCL5), matrix remodelling enzymes (MMP1), angiogenic factors (PDGFA, PDGFB, ANGPT1), and the neuroimmune modulator BDNF exhibited strong correlations between the SomaScan and NULISA platforms, as shown in Fig. 3 . Plasma Proteomic Changes Linked to NSCLC Tumour Recurrence Comparative analysis of feature selection performed on NULISA and SomaScan data independently using Stabl and ALasso methods showed recurrence-related changes in plasma proteins that only reached statistical significance when using ALasso selection on pre-surgical samples (P = 0.015, AUROC = 0.85) on the SomaScan panel (Fig. 4 A). Higher expression of mitochondrial cytochrome c oxidase subunit 7A2-like (COX7A2L), fibroblast growth factor 19 (FGF19), and Testican-2 (SPOCK2) was seen in NSCLC patients with tumour recurrence. The low-affinity immunoglobulin epsilon Fc receptor (FCER2), Fc receptor-like A (FCRLA) and SLIT and NTRK-like protein 2 (SLITRK2) had lower expression in recurrent versus non-recurrent cases (Fig. 4 B, D). Although Stabl did not demonstrate strong selection power (P = 0.062, AUROC = 0.77), FCER2, SPOCK2 and COX7A2L markers were similarly selected by this method (Fig. 4 C), and therefore we sought to use these features to probe for pathway enrichment associated with these markers. Pathway enrichment analysis revealed significant involvement of the identified markers in lymph node development (P = 0.0013), positive regulation of lymphocytes (P = 0.00097), cytokine–cytokine receptor interaction (P = 4.27e-10), and TNF receptor superfamily signaling (P = 0.0016). Plasma Proteomic Signatures pre/post immunotherapy To identify plasma protein markers associated with immunotherapy response, we profiled 30 plasma samples collected from NSCLC patients before initiation of immunotherapy using the NULISAseq assay. Patients were categorised as responders (CR, PR) or non-responders (SD, PD) based on their RECIST criteria. Associated proteins were identified through the two feature selection methods applied separately to data from each assay, which yielded five markers with strong selection signals and statistically significant P-values (P = 0.012, AUROC = 0.83) (Fig. 5 A). Among those, Interleukin-2 receptor subunit alpha (IL-2RA), Interleukin-6 (IL-6), C-reactive protein (CRP), C-C motif chemokine 7 (CCL7) and C-C motif chemokine 19 (CCL19) are well-recognised for their roles in pro-inflammatory and immune activation processes [ 18 ]. Cell surface glycoprotein CD200 receptor 1 (CD200R1) is associated with immune regulation, Osteopontin (SPP1) is typically produced in response to inflammation, and programmed cell death protein 1 (PDCD1), C-type lectin domain family 4 member A (CLEC4A) and Leukaemia inhibitory factor (LIF) are immunosuppressive markers [ 19 ]. Univariate analysis and differential expression box plots showed that all highlighted markers were downregulated in responders except CLEC4A (Fig. 5 B, C). According to KEGG, STRING, and Reactome databases, these markers are significantly enriched in pathways related to cell migration (P = 0.0032), inflammatory response (P = 0.00064), chemokine signalling (P = 5.80e-05), and IL-6 family signalling (P = 0.042), highlighting both immunological and metabolic mechanisms. A subset of 11 patients with both baseline and follow-up plasma samples, collected 3–6 months (median follow-up time: 4.5 months) after immunotherapy, were analysed. Despite achieving a high AUROC, neither analysis model reached statistical significance due to the small sample size (P = 0.12, AUROC = 0.80). Six markers of treatment outcome, including C-C motif chemokine 21 (CCL21), CD276 antigen (B7-H3), IL-2RA, and Interleukin-24 (IL-24), Interleukin 1 receptor-like 1 (IL-1RL1) and Receptor for Advanced Glycation End-products or AGER, were identified by both models (Fig. 5 D). Survival analysis Survival analysis, using a regularised Cox Proportional Hazards model on proteins selected by ALasso, of patients who received immunotherapy, identified nine differentially expressed markers (concordance index = 0.54) (Fig. 6 A). Stabl provided validation of this selection, by reducing this to four key proteins (concordance index = 0.80) when performing feature selection using a CoxPH fit to the censored progression-free-survival time (Fig. 6 C). These four proteins were the cell adhesion molecule CEACAM5 (Hazard Ratio (HR) = 0.18), pentraxin-related protein PTX3 (PTX3) (HR = 0.065), fibroblast growth factor 23 (FGF23) (HR = 0.077), and amphiregulin (AREG) (HR = 0.31). Their increased expression negatively correlated with patient survival. We used these four selected proteins from the survival analysis in an unweighted logistic regression model fit to predict grouped response (CR/PR and SD/PD RECIST scores) for all baseline IO patients, achieving AUROC scores of 0.763 in k-fold validation tests, but with the model not reaching statistical significance (P = 0.073) (Fig. 6 B). Reactome pathway analysis and Pfam domain analysis further highlighted enriched proteins associated with MAPK1/MAPK3 signalling (P = 0.0197) and the interleukin-6 receptor alpha chain (P = 0.0241), suggesting potential immunoregulatory mechanisms. Confounder analysis. We evaluated the impact of clinical variables on the model's performance, including PD-L1 tumour proportion score (TPS), overall survival, disease stage, as well as different treatment modalities, such as chemotherapy, immunotherapy, and targeted therapy. Among these variables, the disease stage emerged as the only significant confounding factor, demonstrating a notable influence on the model’s performance in predicting patient outcomes for those who received immunotherapy (P = 0.047) (Supplementary Table 1). This finding underscores the importance of considering disease stage as a critical factor when assessing the prognostic value of biomarkers in immunotherapy-treated patients. Discussion Blood proteomic biomarkers offer a real-time and non-invasive way to capture dynamic tumour and immune-related signals to better guide disease stratification and therapeutic outcomes for surgery and targeted therapies. We profiled the plasma proteome of 56 NSCLC patients at 87 timepoints, covering various clinical modalities including before and after lung resection surgery, recurrent versus non-recurrent post-surgery, and pre- versus post-immunotherapy. Our analysis used two plasma proteomics platforms, NULISAseq (Alamar Biosciences) and SomaScan (SomaLogic). By applying false discovery-corrected feature-selection methods, we identified distinct protein signatures associated with different clinical groups, including surgical, recurrence, and ICI therapy. NULISAseq employs a highly sensitive dual-antibody sandwich approach with DNA barcoding and NGS, achieving attomolar sensitivity by significantly reducing background signal. Being highly sensitive enables this platform to detect low-abundance proteins like cytokines and immune regulators using a smaller panel of 250 probes. In contrast, SomaScan uses aptamer technology, where numerous oligonucleotide probes cover 7596 proteins with high throughput. These differences can lead to variability in protein detection between the two platforms, where aptamer and antibody-based assays may indicate variable correlations depending on target characteristics [ 20 ]. SomaScan's broader coverage facilitated the detection of a diverse array of proteins, including those involved in extracellular matrix remodeling and metabolic regulation. Our direct comparison of these platforms highlights their complementary strengths and specific instances where their detection capabilities vary. The notable cross-platform correlation emphasizes the robustness and reproducibility of our findings, pointing to that both platforms effectively capture comparable biological variations, despite differences in technology, scale of measurements, and panel focus. In overlapping regions of the two protein panels, correlation of essential immune, vascular, and tissue remodeling markers were found across the two independent proteomic platforms. These include chemokines that facilitate immune cell recruitment (CCL28, CCL17, CXCL2, CCL5), as well as mediators involved in extracellular matrix remodeling (MMP1) and angiogenesis (PDGFA, PDGFB, ANGPT1). These variations have noticeable implications for biomarker discovery and interpretation, emphasising the importance of understanding platform-specific characteristics when selecting biomarkers for translational studies. Detectable biomarkers in the blood, such as those involved in inflammation, coagulation factors, and stress hormones, varied between pre-surgery and post-surgery, provide insights into the biological processes associated with surgical intervention. We observed a reduction in IL-36G in post-surgery samples, a pro-inflammatory cytokine known for activating CD8 + T cells and NK cells, which may reflect a dampening of systemic inflammation following tumour removal [ 21 ]. Conversely, we found elevated circulating levels of MUC16 after surgery, a glycoprotein expressed on the surface of various epithelial cells, reported to be higher in different types of lung cancer, particularly in advanced stages [ 22 , 23 ]. We observed, post-surgery downregulation of CLEC3B, KXD1 and TGFBI alongside upregulation of SHC1 and THBS3 [ 24 ], with pathway analysis results further supporting the critical role of cytokines and extracellular matrix components in distinguishing pre-surgery from post-surgery conditions. Reduction of CLEC3B and TGFBI, both associated with tumour ECM remodelling and tissue homeostasis), together with reduced KXD1, linked to lysosomal function, may reflect early systemic responses to surgical intervention. In parallel, increased SHC1 and THBS3, which are involved in wound healing, angiogenesis, and ECM remodelling, likely indicate active tissue repair and remodelling processes [ 25 ]. These expression changes are consistent with biological adaptations following surgery and may reflect wound healing and tissue reconstruction dynamics. NSCLC recurrence rate following surgery remains a clinical challenge, with rates ranging from 15–20% [ 26 ]. Improving postoperative monitoring and treatment strategies depends on the identification of recurrence-associated biomarkers [ 27 ]. We found biomarkers involved in immune regulation and inflammation (FCER2, FCRLA and SLITRK2), metabolic adaptation (COX7A2L), cell signalling and growth regulation (FGF19), and extracellular matrix regulation (SPOCK2) associated with tumour recurrence. While no study to date has specifically explored the plasma levels of FCER2, FCRLA and SLITRK2 in lung cancer following surgery, their higher expression within tumour tissue has been linked to better overall survival, reduced risk of metastasis, and more favourable outcomes. In contrast, we found the opposite trend, with patients who experienced recurrence after surgery having lower levels of these markers compared to those who remained cancer-free. Since COX7A2L participates in the stabilisation of mitochondrial respiratory supercomplexes, it is expected to decrease in the hypoxic environment of the TME, as reported by Hollinshead et al., who observed reduced mRNA and protein levels in lung cancer cells [ 28 ]. Although the role of circulating COX7A2L in post-surgery lung cancer recurrence has not been specifically investigated, we observed higher plasma levels of this marker in patients who experienced tumour recurrence compared to those who remained disease-free. A similar opposite trend between previous tissue studies and our plasma profiling in recurrent cases was also seen for SPOCK2 [ 29 ]. In contrast, and in line with our plasma findings, Chen et al. reported elevated tissue expression of FGF19 as an independent prognostic indicator for recurrence following NSCLC surgical resection [ 30 ]. Immunotherapy has transformed lung cancer care, yet predictors of treatment response remain intangible. We profiled baseline plasma samples from patients prior to ICI therapy and categorised biomarkers according to RECIST-defined responses. Among the markers that were elevated in non-responders (SD/PD) relative to responders (CR/PR), there is supporting prior publications linking IL-6 [ 31 ], IL-2RA [ 32 ], SPP1 [ 33 ], CD200R1 and CCL19 to systemic inflammation, immune suppression, and/or poor outcomes. IL-6 is reported to have contrasting effects on immunotherapy. In one cohort of 125 advanced NSCLC patients, low baseline IL-6 (< 13.1 pg/mL) was associated with significantly higher objective response and disease control rates and longer PFS/OS under anti-PD-1/PD-L1 therapy [ 34 ], in line with our findings. On the other hand, another study found that patients with high IL-6 had poor outcomes (median PFS ~ 1.9 vs 6.3 months) [ 35 ]. Our finding that circulating IL-2RA is elevated in non-responders confirms the results of a previous study that found high baseline IL-2RA serum levels predicted poor PD-1/PD-L1 therapy response [ 32 ]. SSP1 is a pleiotropic secreted protein overexpressed in lung cancer, which promotes tumour growth, metastasis and creates an immunosuppressive microenvironment [ 33 ]. We found that higher baseline SPP1 levels were linked to worse ICI response, which is supported by other studies showing high pre-treatment serum SPP1 correlates with poor clinical response and higher mortality in NSCLC patients treated with nivolumab and pembrolizumab [ 33 ]. In addition to these known markers, we identified CD200R1 and CCL19 as unique immunotherapy response markers not previously reported in plasma studies. CD200R1 is a receptor on myeloid/immune cells that transmits inhibitory signals from the CD200 ligand. CCL19 attracts CCR7 + T cells and dendritic cells, with higher expression in the TME correlating with enhanced immune infiltration and better clinical outcomes [ 36 , 37 ]. Their opposing immunological roles point to the complex regulatory balance and underexplored pathways within the TME that shape the response to therapy and influence outcomes. To extend our findings beyond baseline predictors, we examined circulating markers in 11 patients with matched pre- and post-ICI therapy plasma samples. We identified six markers: CCL21, CD276, IL-2RA, IL-24, IL-1RL1 and AGER, with only IL-2RA having been previously linked to treatment outcome. CCL21, a chemokine involved in lymphocyte trafficking and dendritic cell homing, has not been reported in plasma-based lung cancer studies, though higher plasma levels have been observed in melanoma patients who responded to immunotherapy. CD276, considered an immune checkpoint molecule, was differentially expressed before and after treatment. Torres-Martínez et al . reported significantly higher levels of soluble CD276 in baseline samples of non-responding NSCLC patients by ELISA [ 38 ]. IL-24, which has immune-modulatory and tumour-suppressive characteristics, was also differentially expressed, with responders in our cohort showing higher baseline IL-24 compared to non-responders. To our knowledge, no study has compared plasma IL-24 levels before and after immunotherapy. However, Zhang et al . found that IL-24 improved prognosis in NSCLC patients, with serum IL-24 levels negatively correlating with TNM classification, consistent with a tumour-suppressive role in NSCLC [ 39 ]. The plasma level of IL-1RL1 was found to have increased at the time of relapse in NSCLC patients with cancer-associated cachexia (CAC) compared with the non-CAC group, showing a negative correlation of this marker within the circulation with the tumour presence [ 40 ]. Its reduction in the follow-up samples upon immunotherapy in our cohort is also proving the same trend for this protein. Finally, we assessed AGER, encoding the Receptor for Advanced Glycation End-products (RAGE), a multiligand pattern-recognition receptor abundantly expressed in lung alveolar cells and involved in inflammation and tissue stress responses. Multiple studies have reported that soluble RAGE negatively correlates with lung cancer, with serum levels decreasing during disease progression [ 41 ]. Although we observed a similar trend, none of these studies investigated the impact of immunotherapy on circulating RAGE levels. Existing predictors of response, such as PD-L1 expression, are inconsistent [ 10 ]. By employing a Cox proportional-hazards model on pretreatment plasma proteomes, alongside the application of the Stabl framework to reduce false discoveries from high-plex panels, we identified a four-protein signature for overall survival: CEACAM5, PTX3, FGF23, and AREG. Notably, higher baseline levels of all these proteins predicted worse outcomes in our cohort. CEACAM5, a member of the carcinoembryonic antigen (CEA) family, acts as an oncofetal growth factor and cell-adhesion molecule. Multiple studies have connected higher preoperative or pretreatment CEA levels with worse prognosis in NSCLC [ 42 ]. Dall'Olio et al . further showed that a post-treatment drop in CEA was associated with longer survival under PD-1/PD-L1 therapy [ 43 ]. Although some reports differ on how circulating CEA relates to therapy response [ 44 ], most available evidence aligns with our findings, showing that higher baseline CEA correlates with greater risk of poor survival [ 45 ]. PTX3 (Pentraxin-3) is involved in inflammation and innate immunity, released by myeloid and stromal cells in response to inflammatory cytokines, where it modulates angiogenesis, complement activation, and tissue remodeling. Elevated PTX3 levels, in both tumour tissue and circulation, predict worse overall survival across several malignancies [ 46 ]. While PTX3 has not been extensively studied in lung-cancer immunotherapy, Diamandis et al . recently found that plasma PTX3 may differentiate lung cancer patients from high-risk smokers with performance comparable to current lung markers [ 47 ]. This aligns with our observation that elevated baseline PTX3 predicts poor survival, possibly reflecting a pro-tumour inflammatory environment. FGF23 has not been reported in prior NSCLC plasma proteome studies, nor has its association with survival upon immunotherapy been explored. Although not considered a "classic" lung cancer marker, higher circulating FGF23 has been linked to shorter overall survival in patients with cancers involving bone metastases [ 48 ]. Finally, AREG, an EGF-family ligand and autocrine growth factor that binds EGFR to drive epithelial proliferation, was selected by our model, with overexpression correlating with shorter overall survival in NSCLC patients. AREG is secreted by lung tumour-infiltrating dendritic cells in the tumour microenvironment, promoting cancer cell growth and metastasis [ 49 ]. From pre- and post-surgery comparisons to recurrence risk, immunotherapy response, and survival outcomes, our plasma proteomics analysis identified significant biological patterns across the examined clinical modalities, particularly immune-related markers. IL-6, IL-2RA, LAG3, CD276, CCL19, and CCL21 were among the immune-regulatory proteins most frequently selected across models, indicating a broad role of systemic immune regulation in both treatment response and disease progression. In various instances, poorer outcomes were associated with markers such as SPP1 and AREG, which are known to influence immune suppression and tumour–stromal interactions. Strong predictors of survival also included metabolic or checkpoint-related markers like FGF23 and CEACAM5, along with inflammatory mediators such as PTX3. In comparisons between surgery and recurrence, tissue remodeling and extracellular matrix-associated proteins, including THBS3 and CLEC3B, were more prominent, suggesting a potential role for them in micrometastatic activity or tumour resection response. Furthermore, all markers were selected using two statistical models, adaptive Lasso and Stabl, which enhances the likelihood that they are relevant despite the large feature space induced by these high-plex panels. We demonstrated how employing two independent high-dimensional technologies can uncover complementary and therapeutically significant plasma signals by combining Alamar's NULISAseq platform's ultra-sensitive quantification of low-abundance proteins with SomaScan's extensive proteome coverage. In addition to emphasising immunological and stromal pathways that are common across lung cancer stages, these results collectively provide a foundation for developing reliable blood-based biomarker panels for treatment stratification and personalised therapy. This study has some limitations to take into consideration. The small sample size may limit the generalizability of these results, particularly when considering recurrence or treatment response groups. Although we utilised two complimentary proteomics technologies, each has advantages and disadvantages, and variations in target coverage or sensitivity may affect the outcomes, especially when comparing platforms. For instance, the NULISA panel only partially overlaps with the more general SomaScan panel since it mainly focuses on proteins linked to inflammation. Additionally, we studied plasma samples from only two timepoints, before and after surgery and immunotherapy, so we can only get an overview of the changes in protein levels over time. Furthermore, we are unable to directly connect the changes in plasma proteins to the processes occurring within the tumour or its microenvironment as we did not incorporate corresponding tissue analyses. Lastly, even with the use of well-established normalization techniques, we may not have been able to completely control for variables like sample processing, systemic inflammation, or other medical conditions that could affect plasma protein levels. Conclusion In summary, our study has identified a number of notable biomarkers using a plasma proteomics screen of various NSCLC clinical groups leveraging high-plex and highly-sensitive blood proteomics data and a comprehensive data analysis pipeline. Declarations Conflicts of Interest JH is a director and GM an employee of SurgeCare, SAS. JM is an employee of Standard Biotools. AK is on the Scientific Advisory Board for Omapix Solutions, Predxbio, Molecular Instruments, and Visiopharm. Competing Interests JH is a director and GM an employee of SurgeCare, SAS. JM is an employee of Standard Biotools. AK is on the Scientific Advisory Board for Omapix Solutions, Predxbio, Molecular Instruments, and Visiopharm. Author Contribution Concept: MA, MEW, AK Experimentation: VYN, SD, CO, WM, JM1, KOB Data analysis: CYN, AK1, CL, JH, GM Writing and critical review: all authors. *AK=Arutha Kulasinghe AK1= Aaron Kilgallon, JM=James Mansfield, JM1 = James Monkman Acknowledgement This study was supported by the Frazer Institute (University of Queensland), Queensland Spatial Biology Centre (Wesley Research Institute), Cure Cancer and The PA Research Foundation. The authors acknowledge the Alamar Biosciences and Somalogic technology access programs. Data Availability All relevant experimental data are available from the corresponding authors upon reasonable request. References Siegel, R.L., K.D. Miller, and A. Jemal, Cancer statistics, 2020 . CA Cancer J Clin, 2020. 70(1): p. 7–30. Zhou, J., et al., Global burden of lung cancer in 2022 and projections to 2050: Incidence and mortality estimates from GLOBOCAN . Cancer Epidemiology, 2024. 93: p. 102693. Crosby, D., et al., Early detection of cancer . Science, 2022. 375(6586): p. eaay9040. Albanes, D., et al., The blood proteome of imminent lung cancer diagnosis . Nature Communications, 2023. 14(1): p. 3042. Rathod, R., et al., Factors influencing false positive low-dose computed tomography-based lung cancer screening in an urban, safety-net system . Cancer Research, 2025. 85(8_Supplement_1): p. 7408–7408. Patz, E.F., et al., Overdiagnosis in low-dose computed tomography screening for lung cancer . JAMA internal medicine, 2014. 174(2): p. 269–274. Mehan, M.R., et al., Protein Signature of Lung Cancer Tissues . PLOS ONE, 2012. 7(4): p. e35157. Macklin, A., S. Khan, and T. Kislinger, Recent advances in mass spectrometry based clinical proteomics: applications to cancer research . Clinical Proteomics, 2020. 17(1): p. 17. Zhao, R., et al., Plasma proteomic profiles for early detection and risk stratification of non-small cell lung carcinoma: A prospective cohort study with 52,913 participants . International Journal of Cancer, 2025. Bar, J., et al., Biological insights from plasma proteomics of non-small cell lung cancer patients treated with immunotherapy . Frontiers in Immunology, 2024. Volume 15–2024. Tognetti, M., et al., Biomarker Candidates for Tumors Identified from Deep-Profiled Plasma Stem Predominantly from the Low Abundant Area . J Proteome Res, 2022. 21(7): p. 1718–1735. Geyer, P.E., et al., The Circulating ProteomeTechnological Developments, Current Challenges, and Future Trends . Journal of Proteome Research, 2024. 23(12): p. 5279–5295. Yaghoubi Naei, V., et al., Advances in novel strategies for isolation, characterization, and analysis of CTCs and ctDNA . Therapeutic Advances in Medical Oncology, 2023. 15: p. 17588359231192401. Hédou, J., et al., Discovery of sparse, reliable omic biomarkers with Stabl . Nature Biotechnology, 2024. 42(10): p. 1581–1593. Hodi, F.S., et al., Evaluation of Immune-Related Response Criteria and RECIST v1.1 in Patients With Advanced Melanoma Treated With Pembrolizumab . J Clin Oncol, 2016. 34(13): p. 1510–7. Meinshausen, N. and P. Bühlmann, Stability selection . Journal of the Royal Statistical Society Series B: Statistical Methodology, 2010. 72(4): p. 417–473. Zou, H., The adaptive lasso and its oracle properties . Journal of the American statistical association, 2006. 101(476): p. 1418–1429. Wu, W., et al., IL-2Rα-biased agonist enhances antitumor immunity by invigorating tumor-infiltrating CD25 + CD8 + T cells . Nature Cancer, 2023. 4(9): p. 1309–1325. Lin, X., et al., Regulatory mechanisms of PD-1/PD-L1 in cancers . Molecular Cancer, 2024. 23(1): p. 108. Raffield, L.M., et al., Comparison of proteomic assessment methods in multiple cohort studies . Proteomics, 2020. 20(12): p. 1900278. Baker, K.J., et al., IL-36 expression is increased in NSCLC with IL-36 stimulation of lung cancer cells promoting a pro-tumorigenic phenotype . Cytokine, 2023. 165: p. 156170. Kimura, Y., et al., Serum CA125 level is a good prognostic indicator in lung cancer . British Journal of Cancer, 1990. 62(4): p. 676–678. Nakamura, H. and T. Nishimura, History, molecular features, and clinical importance of conventional serum biomarkers in lung cancer . Surgery Today, 2017. 47(9): p. 1037–1059. Ly, D., et al., Role for High-Affinity IgE Receptor in Prognosis of Lung Adenocarcinoma Patients . Cancer Immunology Research, 2017. 5(9): p. 821–829. Liang, Y., et al., The increased expression and aberrant methylation of SHC1 in non–small cell lung cancer: Integrative analysis of clinical and bioinformatics databases . Journal of Cellular and Molecular Medicine, 2021. 25(14): p. 7039–7051. Bertoglio, P., et al., EP07.04-19 Clinical and Molecular Profile and Risk Factors of Stage IA NSCLC Relapsing after Radical Resection . Journal of Thoracic Oncology, 2023. 18(11): p. S552-S553. Schuurbiers, M.M.F., et al., Recurrence prediction using circulating tumor DNA in patients with early-stage non-small cell lung cancer after treatment with curative intent: A retrospective validation study . PLOS Medicine, 2025. 22(4): p. e1004574. Hollinshead, K.E., et al., Respiratory supercomplexes promote mitochondrial efficiency and growth in severely hypoxic pancreatic cancer . Cell reports, 2020. 33(1). Wang, Y., et al., Nebulized inhalation of extracellular vesicles containing SPOCK2 suppresses lung adenocarcinoma progression via MAPK inhibition . Discover Oncology, 2025. 16(1): p. 1–14. Chen, J., et al., Enhanced expression of FGF19 predicts poor prognosis in patients with non-small cell lung cancer . Journal of thoracic disease, 2021. 13(3): p. 1769. Naqash, A.R., et al., Increased interleukin-6/C-reactive protein levels are associated with the upregulation of the adenosine pathway and serve as potential markers of therapeutic resistance to immune checkpoint inhibitor-based therapies in non-small cell lung cancer . J Immunother Cancer, 2023. 11(10). Siemiątkowska, A., et al., Serum sCD25 Protein as a Predictor of Lack of Long-Term Benefits from Immunotherapy in Non-Small Cell Lung Cancer: A Pilot Study. Cancers (Basel), 2021. 13(15). Ramoni, D., et al., Circulating Osteopontin Predicts Clinical and Radiological Response in First-Line Treatment of Advanced Non-Small Cell Lung Cancer . Lung, 2024. 202(2): p. 197–210. Kang, D.H., et al., Baseline Serum Interleukin-6 Levels Predict the Response of Patients with Advanced Non-small Cell Lung Cancer to PD-1/PD-L1 Inhibitors . Immune Netw, 2020. 20(3): p. e27. Su, S., et al., Predictive value of peripheral blood biomarkers in patients with non-small-cell lung cancer responding to anti-PD-1-based treatment . Cancer Immunology, Immunotherapy, 2024. 73(1): p. 12. Gu, Q., et al., CCL19: a novel prognostic chemokine modulates the tumor immune microenvironment and outcomes of cancers . Aging (Albany NY), 2023. 15(21): p. 12369–12387. Shao, A. and D.M. Owens, The immunoregulatory protein CD200 as a potentially lucrative yet elusive target for cancer therapy . Oncotarget, 2023. 14(1). Torres-Martínez, S., et al., Circulating Immune Proteins: Improving the Diagnosis and Clinical Outcome in Advanced Non-Small Cell Lung Cancer . International Journal of Molecular Sciences, 2023. 24(24): p. 17587. Zhang, K., et al., Serum IL-24 combined with CA125 as screening and prognostic biomarkers for NSCLC . Cell Biology International, 2024. 48(8): p. 1160–1168. Al-Sawaf, O., et al., Body composition and lung cancer-associated cachexia in TRACERx . Nature Medicine, 2023. 29(4): p. 846–858. Jing, R., et al., Receptor for advanced glycation end products (RAGE) soluble form (sRAGE): a new biomarker for lung cancer . Neoplasma, 2010. 57(1): p. 55. Duan, F., et al., A novel diagnostic model for predicting immune microenvironment subclass based on costimulatory molecules in lung squamous carcinoma . Frontiers in Genetics, 2022. Volume 13–2022. Dall'Olio, F.G., et al., CEA and CYFRA 21 – 1 as prognostic biomarker and as a tool for treatment monitoring in advanced NSCLC treated with immune checkpoint inhibitors . Ther Adv Med Oncol, 2020. 12: p. 1758835920952994. Dal Bello, M., et al., The role of CEA, CYFRA21-1 and NSE in monitoring tumor response to Nivolumab in advanced non-small cell lung cancer (NSCLC) patients . Journal of Translational Medicine, 2019. 17: p. 1–10. Marin-Acevedo, J.A., E.O. Kimbrough, and Y. Lou, Next generation of immune checkpoint inhibitors and beyond . Journal of Hematology & Oncology, 2021. 14(1): p. 45. Jung, H., et al., Prognostic Value of Pentraxin3 Protein Expression in Human Malignancies: A Systematic Review and Meta-Analysis . Cancers (Basel), 2024. 16(22). Diamandis, E.P., et al., Pentraxin-3 is a novel biomarker of lung carcinoma . Clin Cancer Res, 2011. 17(8): p. 2395–9. Mansinho, A., et al., Levels of Circulating Fibroblast Growth Factor 23 (FGF23) and Prognosis in Cancer Patients with Bone Metastases . International Journal of Molecular Sciences, 2019. 20(3): p. 695. Hsu, Y.L., et al., Lung tumor-associated dendritic cell-derived amphiregulin increased cancer progression . J Immunol, 2011. 187(4): p. 1733–44. Additional Declarations Competing interest reported. JH is a director and GM an employee of SurgeCare, SAS. JM is an employee of Standard Biotools. AK is on the Scientific Advisory Board for Omapix Solutions, Predxbio, Molecular Instruments, and Visiopharm. Supplementary Files graphicalabstract.png Graphical abstract. Evolution of plasma proteomics technologies and translational applications of plasma proteomics. TableS1.docx Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 11 Jan, 2026 Reviews received at journal 24 Nov, 2025 Reviews received at journal 07 Oct, 2025 Reviewers agreed at journal 06 Oct, 2025 Reviewers agreed at journal 22 Sep, 2025 Reviewers invited by journal 29 Jul, 2025 Editor assigned by journal 19 Jul, 2025 Submission checks completed at journal 17 Jul, 2025 First submitted to journal 14 Jul, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7118825","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":492449347,"identity":"6c62a2c6-ef68-41f5-9171-6c2544d42117","order_by":0,"name":"Vahid Yaghoubi Naei","email":"","orcid":"","institution":"University of Technology Sydney","correspondingAuthor":false,"prefix":"","firstName":"Vahid","middleName":"Yaghoubi","lastName":"Naei","suffix":""},{"id":492449349,"identity":"82fea403-238c-49c2-9488-d280b5ee4d09","order_by":1,"name":"Aaron Kilgallon","email":"","orcid":"","institution":"The Wesley Hospital","correspondingAuthor":false,"prefix":"","firstName":"Aaron","middleName":"","lastName":"Kilgallon","suffix":""},{"id":492449350,"identity":"5457904a-7d11-49da-b560-b69cc326bf3f","order_by":2,"name":"Gwendoline Mendes","email":"","orcid":"","institution":"SurgeCare","correspondingAuthor":false,"prefix":"","firstName":"Gwendoline","middleName":"","lastName":"Mendes","suffix":""},{"id":492449351,"identity":"9dfb6150-0b15-4a46-be9b-95af51e83594","order_by":3,"name":"Sanjay Dutta","email":"","orcid":"","institution":"The Princess Alexandra Hospital","correspondingAuthor":false,"prefix":"","firstName":"Sanjay","middleName":"","lastName":"Dutta","suffix":""},{"id":492449352,"identity":"770f976b-5f36-403e-b98b-ecad226cbf65","order_by":4,"name":"Clara Lawler","email":"","orcid":"","institution":"University of Queensland","correspondingAuthor":false,"prefix":"","firstName":"Clara","middleName":"","lastName":"Lawler","suffix":""},{"id":492449353,"identity":"293c752b-ab07-4a3d-b8b5-276b782e43a4","order_by":5,"name":"Connor O’Leary","email":"","orcid":"","institution":"The Princess Alexandra Hospital","correspondingAuthor":false,"prefix":"","firstName":"Connor","middleName":"","lastName":"O’Leary","suffix":""},{"id":492449354,"identity":"c97ffd8c-7567-4b29-bd11-f9461686093c","order_by":6,"name":"William Mullally","email":"","orcid":"","institution":"The Princess Alexandra Hospital","correspondingAuthor":false,"prefix":"","firstName":"William","middleName":"","lastName":"Mullally","suffix":""},{"id":492449355,"identity":"702687fc-d06e-4b51-a328-97b2ea36db24","order_by":7,"name":"James Monkman","email":"","orcid":"","institution":"University of Queensland","correspondingAuthor":false,"prefix":"","firstName":"James","middleName":"","lastName":"Monkman","suffix":""},{"id":492449356,"identity":"e6823ab6-d938-4648-8583-3f3d07e807a9","order_by":8,"name":"James Mansfield","email":"","orcid":"","institution":"Standard BioTools Canada Inc","correspondingAuthor":false,"prefix":"","firstName":"James","middleName":"","lastName":"Mansfield","suffix":""},{"id":492449357,"identity":"a2100b9c-f44b-45e8-8724-d5d2bf6521d2","order_by":9,"name":"Julien Hedou","email":"","orcid":"","institution":"Sorbonne Université, INSERM, Centre de Recherche Saint-Antoine, CRSA","correspondingAuthor":false,"prefix":"","firstName":"Julien","middleName":"","lastName":"Hedou","suffix":""},{"id":492449358,"identity":"bbc236b8-586a-44e6-a36b-a90e85bd4885","order_by":10,"name":"Mark N. Adams","email":"","orcid":"","institution":"Queensland University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Mark","middleName":"N.","lastName":"Adams","suffix":""},{"id":492449359,"identity":"78e0dc65-a693-4647-a934-8553585cc0a2","order_by":11,"name":"Ken O’Byrne","email":"","orcid":"","institution":"The Princess Alexandra Hospital","correspondingAuthor":false,"prefix":"","firstName":"Ken","middleName":"","lastName":"O’Byrne","suffix":""},{"id":492449360,"identity":"19bccb9c-3b60-47a4-a5c9-339de4e882a8","order_by":12,"name":"Majid E Warkiani","email":"","orcid":"","institution":"University of Technology Sydney","correspondingAuthor":false,"prefix":"","firstName":"Majid","middleName":"E","lastName":"Warkiani","suffix":""},{"id":492449361,"identity":"f34d42fc-c4b6-4b91-bdb7-38a63a2b5924","order_by":13,"name":"Arutha Kulasinghe","email":"data:image/png;base64,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","orcid":"","institution":"University of Queensland","correspondingAuthor":true,"prefix":"","firstName":"Arutha","middleName":"","lastName":"Kulasinghe","suffix":""}],"badges":[],"createdAt":"2025-07-14 08:38:22","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7118825/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7118825/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":88095548,"identity":"377cdfbe-1a18-4390-9789-09376d7f041f","added_by":"auto","created_at":"2025-08-01 10:54:32","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":2051748,"visible":true,"origin":"","legend":"\u003cp\u003ePatient clinical information chart showing age, sex, treatment mode, stage of the disease, PDL1 tumour proportion score, histology subtype and best overall response to therapy (top). Comparison of 20 lung cancer patients at two time points (pre- and post-resection surgery), cured vs. recurrent, using the SomaScan assay. Comparison of the plasma profiles of 25 baseline immunotherapy patients and pre- and post-immunotherapy plasma profiles for 11 patients using the NULISAseq assay (bottom). N=number of patients\u003c/p\u003e","description":"","filename":"figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-7118825/v1/f60b4b061f7c44dc77593e6e.png"},{"id":88095554,"identity":"59e084a4-3015-46d6-b94b-3e10d0db0c7c","added_by":"auto","created_at":"2025-08-01 10:54:32","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":5804358,"visible":true,"origin":"","legend":"\u003cp\u003eComparative analysis of pre- and post-surgery plasma proteomics A) Box plots of Adaptive Lasso model predictions (left) and corresponding ROC curves (right) of selected proteins from the NULISA panel (P = 0.002). B) Expression values of selected proteins of the Adaptive Lasso model using NULISA data. C) Box plots of Stabl+Ridge Regression model predictions (left), corresponding ROC curves (middle), and Stabl regularisation paths (right) for selected proteins from NULISA data (P = 0.01). D) Expression of Stabl-selection proteins using NULISA data. E) Box plots of model predictions (left), corresponding ROC curves (right) for proteins selected by Adaptive Lasso feature selection on SomaScan data (P = 0.001). F) Expression values of selected proteins of the Adaptive Lasso model using the SomaScan data. G) Box plots of Stabl+Logistic Regression model predictions (left), corresponding ROC curves (middle), and Stabl regularisation paths (right) for selected proteins from the SomaScan panel. H) Expressions of selection proteins using SomaScan data. I) Univariate analysis of NULISA protein expression (left) and SomaScan protein expression (right) shows the comparative advantage of the high-sensitivity NULISA assay in detecting significant log-fold changes in protein expression.\u003c/p\u003e","description":"","filename":"figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-7118825/v1/09a93258310ad68736b85f04.png"},{"id":88505280,"identity":"e08af101-714a-4920-802a-78707b917a12","added_by":"auto","created_at":"2025-08-07 07:22:49","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":4166201,"visible":true,"origin":"","legend":"\u003cp\u003eA heatmap illustrating the cross-platform correlation of overlapping proteins identified through SomaScan and NULISA assays. The left panel presents the complete correlation heatmap, showing correlations in global expression patterns across both platforms in response to surgical intervention. The right panel presents a magnified view of the clustered region that includes strongly correlated markers.\u003c/p\u003e","description":"","filename":"figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-7118825/v1/8c6b94560ff3327bb54884ea.png"},{"id":88098792,"identity":"26b9e70d-e1da-4bee-aa90-37a68d82b21a","added_by":"auto","created_at":"2025-08-01 11:10:32","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":4645534,"visible":true,"origin":"","legend":"\u003cp\u003eComparative analysis of post-surgery plasma proteomic profiles of recurrent patients against non-recurrent patients. A) Box plots of model predictions (left) and corresponding ROC curves (right) for proteins selected using Adaptive Lasso (P = 0.015). B) Expression values of proteins selected by the Adaptive Lasso model using SomaScan data. C) Box plots of Stabl+Ridge regression model predictions (left), corresponding ROC curves (middle), and Stabl regularisation paths using SomaScan data (right). D) Log-fold changes in the SomaScan assay show that a number of univariate statistically significant proteins were selected as informative by the Adaptive Lasso selection method.\u003c/p\u003e","description":"","filename":"figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-7118825/v1/e7e7831fb0fca038f6e17f1e.png"},{"id":88096951,"identity":"f4cf5652-6d49-46d4-9071-b2955e994c1d","added_by":"auto","created_at":"2025-08-01 11:02:32","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":3254968,"visible":true,"origin":"","legend":"\u003cp\u003eComparative analysis of baseline immunotherapy plasma proteomic profiles using the NULISA platform to distinguish clinical response groups. A) Box plots of model predictions (left), corresponding ROC curves (middle), and Stabl regularization paths (right) for distinguishing responders vs. non-responders samples using NULISA data (P = 0.012). B) Expressions values of selected proteins of the Stabl with Ridge regression model for distinguishing responders vs. non-responders using NULISA data. C) Univariate analysis of log-fold differences in protein expression of selected proteins from the NULISA data. D) Expression values of selected proteins of the Adaptive Lasso and Stabl models for comparing differential expressions before and after ICI therapy using the NULISA data.\u003c/p\u003e","description":"","filename":"figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-7118825/v1/55c502212d40117bc6c8c973.png"},{"id":88095551,"identity":"5df14222-98c9-44db-b47f-c9c5a71c6673","added_by":"auto","created_at":"2025-08-01 10:54:32","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":1030325,"visible":true,"origin":"","legend":"\u003cp\u003eSurvival analysis of plasma proteomic profiles selected from the NULISA panel using ALasso and Stabl feature selection methods. A) Cox model for markers selected using ALasso to predict survival endpoints. B) ROC curve demonstrating the predictive performance of the Stabl model in selecting 4 key NULISA markers that can model grouped patient response. C) Regularisation paths of NULISA proteins selected by the Stabl model to differentiate survival endpoints.\u003c/p\u003e","description":"","filename":"figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-7118825/v1/b50770219f00887dc68375df.png"},{"id":88376339,"identity":"1c569100-c715-48df-82aa-2cdfc2639c7d","added_by":"auto","created_at":"2025-08-05 21:28:48","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":14745082,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7118825/v1/7636c0e1-3789-4e8f-bad5-c5ffd3234d15.pdf"},{"id":88099855,"identity":"18e18ab8-da1d-4cce-8049-90a53787b83f","added_by":"auto","created_at":"2025-08-01 11:18:32","extension":"png","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":4490106,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGraphical abstract.\u003c/strong\u003e Evolution of plasma proteomics technologies and translational applications of plasma proteomics.\u003c/p\u003e","description":"","filename":"graphicalabstract.png","url":"https://assets-eu.researchsquare.com/files/rs-7118825/v1/f11c71038b5dd9b4233cb632.png"},{"id":88096949,"identity":"6a65db15-a071-4831-802e-e40f7d7cea0d","added_by":"auto","created_at":"2025-08-01 11:02:32","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":25761,"visible":true,"origin":"","legend":"","description":"","filename":"TableS1.docx","url":"https://assets-eu.researchsquare.com/files/rs-7118825/v1/a57db6501a1928fa3e0e1b31.docx"}],"financialInterests":"Competing interest reported. JH is a director and GM an employee of SurgeCare, SAS. JM is an employee of Standard Biotools. AK is on the Scientific Advisory Board for Omapix Solutions, Predxbio, Molecular Instruments, and Visiopharm.","formattedTitle":"Dissecting Non-Small Cell Lung Cancer (NSCLC) with Blood Proteomics - From Surgical to Immunotherapeutic Responses","fulltext":[{"header":"Introduction","content":"\u003cp\u003eLung cancer, in particular non-small cell lung cancer (NSCLC), remains among the leading causes of cancer-related deaths worldwide and is the second most frequently diagnosed cancer [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Detecting lung cancer at an early stage is a critical need to improve therapeutic options and long-term health outcomes [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. However, NSCLC, the most common form of lung cancer, is often identified at advanced stages, where patient prognosis is often poor. Early detection is thus critical to improving survival. Currently, low-dose computed tomography (LDCT) screening of high-risk individuals (like heavy smokers) can identify early-stage lung cancers and has been shown to reduce lung cancer mortality [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. However, LDCT’s impact is limited by significant drawbacks: screening yields a high rate of false positives (especially in older ages) leading to higher costs and patient anxiety [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Additionally, overdiagnosis of indolent lesions and cumulative radiation exposure pose ongoing concerns [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. These limitations, combined with the fact that many at-risk individuals are ineligible or do not undergo screening, indicate the importance of better non-invasive diagnostic tools. Plasma proteomics offers a non-invasive means to identify these changes and assist early diagnosis, prognostication, and response to therapy monitoring [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. The plasma proteome encompasses a wide range of proteins with different concentrations, reflecting the dynamic state of biomarkers and can be a good proxy for determining tumour dynamics [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. A recent study identified a panel of three plasma proteins that can predict NSCLC risk up to 10 years before diagnosis (achieving ≈ 0.90 area under the curve (AUC) in 5-year predictions) [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Likewise, longitudinal proteomic analyses in patients receiving therapy have linked changes in specific plasma proteins to treatment responses. rising levels of soluble PD-1 in the bloodstream during immune checkpoint inhibitor therapy have been shown to correlate with tumour PD-L1 status and improved patient outcomes [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. These advances highlight the advantages of plasma proteomics as a complementary strategy to imaging, with the potential to improve early NSCLC diagnosis and to enable proactive treatment response monitoring for better clinical management [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e].\u003c/p\u003e\u003cp\u003ePrecise and unbiased affinity-based methods for large-scale proteome analysis like SomaScan (SomaLogic, US) and Nucleic Acid-Linked Immuno-Sandwich Assay (NULISA) (Alamar Biosciences, Canada) have been recently developed, demonstrating high sensitivity and specificity for a wide array of proteins across the entire dynamic range in complex fluids, like serum or plasma. These advancements address the limitations of traditional and even complex mass spectrometry technologies in detecting novel disease biomarkers [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. In this study, we aim to investigate the plasma proteomes of NSCLC patients using SomaScan v4.1, which comprises reagents to detect 7596 plasma proteins including cardiovascular, inflammation and immune response, metabolic diseases, oncology, neurology, cytokines, and respiratory disease biomarkers, and Alamar Biosciences NULISAseq Inflammation Panel, which detects 250 biomarkers, before and after surgery, and before and after immune checkpoint immunotherapy, to discover potential prognostic and diagnostic biomarkers. We utilized the feature selection method Stabl in combination with linear modeling to identify protein signatures associated with plasma proteome changes induced by adjuvant surgery, proteins associated with disease recurrence, proteins prognosticative of response to immunotherapy, and proteins associated with patient survival [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e].\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cb\u003ePatient cohort\u003c/b\u003e\u003c/p\u003e\u003cp\u003eEthics approval for this study was obtained from the Metro South Health District Human Research Ethics Committee under the National Health and Medical Research Council guidelines (HREC/11/QPAH/331) to collect samples from the Princess Alexandra Hospital. Patients who were selected for this study were treatment-naive and without prior cancer diagnosis within the last 5 years. This study has been ratified by the Queensland University of Technology Human Research Ethics Committee. Patients were grouped based on treatment modality (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). For 20 patients who underwent tumour resection surgery, peripheral blood samples were collected within one week prior to surgery and again 2–4 weeks post-surgery. All patients in this group were enrolled with signed informed consent and followed prospectively for clinical outcomes over a period of three to five years. For 36 patients who received immunotherapy (Pembrolizumab, Atezolizumab and Durvalumab), or combination therapy including chemotherapy (Carboplatin, Pemetrexed, Cisplatin, Paclitaxel, Gemcitabine and Docetaxel), targeted therapy (Osimertinib and Crizotinib) and immunotherapy, blood samples were collected at baseline (on the day of or the day before the first round of systemic therapy) for 25 patients, and at baseline and 3–6 months after treatment initiation for 11 patients. Patient responses to immunotherapy were assessed based on RECIST 1.1 criteria [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. All participants were recruited from the Princess Alexandra Hospital thoracic and lung cancer clinic between March 2019 and August 2024.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eBlood sample processing\u003c/b\u003e\u003c/p\u003e\u003cp\u003eBlood samples were collected from 56 patients at 87 timepoints in K2EDTA-coated tubes (BD Diagnostics, US), immediately mixed, and transported to the lab at room temperature. Plasma separation was completed within 4 hours of collection by centrifugation at 2200×g for 15 minutes at 4°C with the break off. The samples were then aliquoted and stored at − 80°C prior to blood proteomic profiling.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eBlood Proteomic profiling of NSCLC plasma samples\u003c/b\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eSomaScan (Somalogic, US)\u003c/b\u003e\u003c/p\u003e\u003cp\u003eMatched plasma samples from 20 patients pre- and post-tumour resection (40 samples in total, 55 µL/sample) were protein profiled using the SomaScan v4.1 assay. Plasma was analysed in the SomaLogic Laboratory in Colorado using a biomarker panel with 7596 SOMAmers (7301 have available UniProt IDs). Samples were randomised and run on 96-well plates alongside external control samples (calibrators, quality control (QC), and buffer). Serial dilutions (0.05%, 0.5%, and 20%) were used to optimise the detection of protein targets, with the full panel of SOMAmers present in all dilutions, a design consistently applied across all samples. Raw data were standardised using external control samples to adjust for variability in microarrays and discrepancies both within and between plates. This process included adaptive normalisation by maximum likelihood (ANML) to an external reference, aimed at reducing inter-sample variability. The final SomaScan data were provided in both ANML and non-ANML formats in relative fluorescence units (RFU), and log-transformed (log-2) for the main analysis. The limit of detection (LOD) for each SOMAmer was determined using external buffer samples. Quality control was performed by comparing the median of QC samples on each plate to the reference, and a cross-plate QC check measure (pass/flag) was assigned to each SOMAmer.\u003c/p\u003e\u003cp\u003e\u003cb\u003eAlamar ARGO HT (Alamar Biosciences, Canada)\u003c/b\u003e\u003c/p\u003e\u003cp\u003eThe Alamar ARGO™ HT system (Model #B31), a fully automated platform designed for multiplex proteomic assays, was used to analyze 25 baseline samples collected before immunotherapy, 11 baseline samples and their matched post-immunotherapy counterparts (22 samples in total), and 17 paired samples collected before and after tumour resection surgery, all of which were matched with the SomaScan assay. Plasma was run on the NULISAseq inflammation Panel 250. Briefly, immunocomplexes were formed using paired oligonucleotide-conjugated antibodies, followed by sequential capture steps with oligo-dT and streptavidin beads, ligation of DNA reporters, and barcoding of samples. The resulting libraries, comprising both target- and sample-specific molecular identifiers, were PCR-amplified, purified, and sequenced using the AVITI platform (Element Biosciences). Each run incorporated multiple layers of quality control, including internal controls (exogenous spike-ins) to normalise individual sample counts, pooled plasma controls to assess intra- and inter-plate consistency, and negative controls (assay buffer only) to evaluate background signal and determine the limits of detection (LOD). Data normalisation followed a multi-step process: raw target counts were first divided by the internal control counts for each sample, then normalised to the median of the inter-plate control for each target and scaled by a factor of 10⁴. A pseudo count of + 1 was added prior to log2 transformation. The final output, expressed as NULISAseq Protein Quantification (NPQ) units, reflects log2-normalised protein expression levels. Plate-specific LODs were determined by calculating the mean plus three standard deviations of the untransformed signal from the negative controls, followed by the same normalisation and log transformation steps. To ensure data quality, any sample with internal control values deviating by more than ± 40% from the plate-wide median was flagged for further evaluation. These samples underwent additional assessment using downstream dimensionality reduction techniques, including principal component analysis and heatmap clustering, to determine their suitability for inclusion in the final dataset.\u003c/p\u003e\u003ch2\u003eStatistical analysis\u003c/h2\u003e\u003cp\u003eWe used a multitiered statistical framework, including descriptive statistics, univariate tests, and false discovery-corrected feature selection. For feature selection, we used adaptive Lasso (ALasso) regression within the Stabl algorithm, a method established by SurgeCare [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e], and standard stability selection using ALasso [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. Stabl is a statistical framework that uses artificial feature injection, stability-based feature selection, and false discovery control of a False Discovery Proportion (FDP) estimate to find reliable sets of predictive biomarkers from high-dimensional datasets. The final fits to the selected markers were varied on the relevant sample subset and ranged from unweighted Logistic Regression models to Ridge-Regression models and l1-weighted Cox Proportional Hazards models. In each pseudo experiment, data were randomly subsampled, artificial features were created using random permutation, and false discovery rates of uninformative features were estimated using these artificial features. Cross-validation was done using a robust Monte Carlo 5x5 method, with stratification by patient ID to prevent data leakage and to preserve the integrity of comparisons between matched samples from the same patient. The feature selection and the choice of a stability threshold of Stabl were adjusted by evaluating several parameters, such as the final regression regularisation parameters, artificial feature generation approach, and, where false-discovery control failed, signal-to-noise ratio. For each comparison modality, the most stable features, defined as those exceeding the optimized data-driven selection frequency threshold and for which the associated model achieved statistical significance in predictions (P \u0026lt; 0.05), were retained. The area under the receiver operating characteristic curve (AUROC) was used to evaluate the predictive performance of the models. To account for multiple comparisons in the pathway analysis, p-values were adjusted using an FDR correction. K-fold tests in the survival analysis to predict patient response were conducted with four folds and ten repeats. Correlation analysis was performed on matched plasma samples examined via both the SomaScan and NULISA platforms. Shared markers consistent across both assays were chosen, and the Pearson correlation of expression levels were clustered using the distance metric to assess regions of cross-platform panel correlation (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cb\u003eSurgery-Induced Alterations in the Plasma Proteome\u003c/b\u003e\u003c/p\u003e\u003cp\u003eMatched patient samples pre- and post-surgery revealed measurable changes in protein expression. The predictive models constructed on the selected proteins demonstrated strong performance, with an AUROC of 0.81 for ALasso (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA) and 0.76 for Stabl (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC). The fold change plot generated with the NULISAseq assay highlighted an opposing trend in the expression of two key markers: Mucin-16 (MUC16) was upregulated after surgery, whereas Interleukin-36 gamma (IL36G) was downregulated (P\u0026thinsp;=\u0026thinsp;0.01). These markers were consistently identified by both feature selection methods (ALasso and Stabl) as leading predictors distinguishing pre- and post-surgical states (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB, D). From the SomaScan assay using both selection models (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eE, G), Thrombospondin-3 (THBS3), Tetranectin (CLEC3B), KxDL motif-containing protein 1 (KDXI), SHC-transforming protein 1 (SHC1), and Transforming growth factor-beta-induced protein ig-h3 (TGFBI) were identified as differentially expressed markers. Differential expression box plots indicated a reduction for CLEC3B, TGFBI, and KDXI, but an increase for THBS3 and SHC1 in post-surgery samples (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eF, H). On both platforms, MUC16, IL36G, THBS3, CLEC3B, and TGFBI were chosen by both selection methods. Comparing the NULISA assay with SomaScan protein expression in a univariate analysis revealed that the former had a higher statistical limit for detecting log-fold changes in protein expression, mainly due to its higher sensitivity (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eI). Pathway enrichment analysis of the selected proteins highlighted key biological processes affected by surgery, including cytokine activity (P\u0026thinsp;=\u0026thinsp;0.0093) and signalling (P\u0026thinsp;=\u0026thinsp;0.031), extracellular compartments (P\u0026thinsp;=\u0026thinsp;2.46e-06), and cytokine\u0026ndash;cytokine receptor interaction (P\u0026thinsp;=\u0026thinsp;0.0016).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eCorrelating Protein Signatures Across Platforms\u003c/b\u003e\u003c/p\u003e\u003cp\u003eTo evaluate the consistency of protein measurements across technologies, we performed a cross-platform correlation analysis of matched pre- and post-surgery samples. This approach enabled us to identify overlapping protein signals and assess agreement between the SomaScan and NULISA platforms. Cross-platform correlation analysis of the pre- and post-surgery samples run on both platforms identified a cluster of highly correlated proteins that consistently exhibited similarity, although the correlation between both panels generally remained positive. Amongst them, Chemokines (CCL28, CXCL2, CCL17, CCL5), matrix remodelling enzymes (MMP1), angiogenic factors (PDGFA, PDGFB, ANGPT1), and the neuroimmune modulator BDNF exhibited strong correlations between the SomaScan and NULISA platforms, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003ePlasma Proteomic Changes Linked to NSCLC Tumour Recurrence\u003c/b\u003e\u003c/p\u003e\u003cp\u003eComparative analysis of feature selection performed on NULISA and SomaScan data independently using Stabl and ALasso methods showed recurrence-related changes in plasma proteins that only reached statistical significance when using ALasso selection on pre-surgical samples (P\u0026thinsp;=\u0026thinsp;0.015, AUROC\u0026thinsp;=\u0026thinsp;0.85) on the SomaScan panel (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA). Higher expression of mitochondrial cytochrome c oxidase subunit 7A2-like (COX7A2L), fibroblast growth factor 19 (FGF19), and Testican-2 (SPOCK2) was seen in NSCLC patients with tumour recurrence. The low-affinity immunoglobulin epsilon Fc receptor (FCER2), Fc receptor-like A (FCRLA) and SLIT and NTRK-like protein 2 (SLITRK2) had lower expression in recurrent versus non-recurrent cases (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB, D). Although Stabl did not demonstrate strong selection power (P\u0026thinsp;=\u0026thinsp;0.062, AUROC\u0026thinsp;=\u0026thinsp;0.77), FCER2, SPOCK2 and COX7A2L markers were similarly selected by this method (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eC), and therefore we sought to use these features to probe for pathway enrichment associated with these markers. Pathway enrichment analysis revealed significant involvement of the identified markers in lymph node development (P\u0026thinsp;=\u0026thinsp;0.0013), positive regulation of lymphocytes (P\u0026thinsp;=\u0026thinsp;0.00097), cytokine\u0026ndash;cytokine receptor interaction (P\u0026thinsp;=\u0026thinsp;4.27e-10), and TNF receptor superfamily signaling (P\u0026thinsp;=\u0026thinsp;0.0016).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003ePlasma Proteomic Signatures pre/post immunotherapy\u003c/b\u003e\u003c/p\u003e\u003cp\u003eTo identify plasma protein markers associated with immunotherapy response, we profiled 30 plasma samples collected from NSCLC patients before initiation of immunotherapy using the NULISAseq assay. Patients were categorised as responders (CR, PR) or non-responders (SD, PD) based on their RECIST criteria. Associated proteins were identified through the two feature selection methods applied separately to data from each assay, which yielded five markers with strong selection signals and statistically significant P-values (P\u0026thinsp;=\u0026thinsp;0.012, AUROC\u0026thinsp;=\u0026thinsp;0.83) (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA). Among those, Interleukin-2 receptor subunit alpha (IL-2RA), Interleukin-6 (IL-6), C-reactive protein (CRP), C-C motif chemokine 7 (CCL7) and C-C motif chemokine 19 (CCL19) are well-recognised for their roles in pro-inflammatory and immune activation processes [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Cell surface glycoprotein CD200 receptor 1 (CD200R1) is associated with immune regulation, Osteopontin (SPP1) is typically produced in response to inflammation, and programmed cell death protein 1 (PDCD1), C-type lectin domain family 4 member A (CLEC4A) and Leukaemia inhibitory factor (LIF) are immunosuppressive markers [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. Univariate analysis and differential expression box plots showed that all highlighted markers were downregulated in responders except CLEC4A (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB, C). According to KEGG, STRING, and Reactome databases, these markers are significantly enriched in pathways related to cell migration (P\u0026thinsp;=\u0026thinsp;0.0032), inflammatory response (P\u0026thinsp;=\u0026thinsp;0.00064), chemokine signalling (P\u0026thinsp;=\u0026thinsp;5.80e-05), and IL-6 family signalling (P\u0026thinsp;=\u0026thinsp;0.042), highlighting both immunological and metabolic mechanisms.\u003c/p\u003e\u003cp\u003eA subset of 11 patients with both baseline and follow-up plasma samples, collected 3\u0026ndash;6 months (median follow-up time: 4.5 months) after immunotherapy, were analysed. Despite achieving a high AUROC, neither analysis model reached statistical significance due to the small sample size (P\u0026thinsp;=\u0026thinsp;0.12, AUROC\u0026thinsp;=\u0026thinsp;0.80). Six markers of treatment outcome, including C-C motif chemokine 21 (CCL21), CD276 antigen (B7-H3), IL-2RA, and Interleukin-24 (IL-24), Interleukin 1 receptor-like 1 (IL-1RL1) and Receptor for Advanced Glycation End-products or AGER, were identified by both models (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eD).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eSurvival analysis\u003c/b\u003e\u003c/p\u003e\u003cp\u003eSurvival analysis, using a regularised Cox Proportional Hazards model on proteins selected by ALasso, of patients who received immunotherapy, identified nine differentially expressed markers (concordance index\u0026thinsp;=\u0026thinsp;0.54) (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eA). Stabl provided validation of this selection, by reducing this to four key proteins (concordance index\u0026thinsp;=\u0026thinsp;0.80) when performing feature selection using a CoxPH fit to the censored progression-free-survival time (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eC). These four proteins were the cell adhesion molecule CEACAM5 (Hazard Ratio (HR)\u0026thinsp;=\u0026thinsp;0.18), pentraxin-related protein PTX3 (PTX3) (HR\u0026thinsp;=\u0026thinsp;0.065), fibroblast growth factor 23 (FGF23) (HR\u0026thinsp;=\u0026thinsp;0.077), and amphiregulin (AREG) (HR\u0026thinsp;=\u0026thinsp;0.31). Their increased expression negatively correlated with patient survival. We used these four selected proteins from the survival analysis in an unweighted logistic regression model fit to predict grouped response (CR/PR and SD/PD RECIST scores) for all baseline IO patients, achieving AUROC scores of 0.763 in k-fold validation tests, but with the model not reaching statistical significance (P\u0026thinsp;=\u0026thinsp;0.073) (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eB). Reactome pathway analysis and Pfam domain analysis further highlighted enriched proteins associated with MAPK1/MAPK3 signalling (P\u0026thinsp;=\u0026thinsp;0.0197) and the interleukin-6 receptor alpha chain (P\u0026thinsp;=\u0026thinsp;0.0241), suggesting potential immunoregulatory mechanisms.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eConfounder analysis.\u003c/b\u003e\u003c/p\u003e\u003cp\u003eWe evaluated the impact of clinical variables on the model's performance, including PD-L1 tumour proportion score (TPS), overall survival, disease stage, as well as different treatment modalities, such as chemotherapy, immunotherapy, and targeted therapy. Among these variables, the disease stage emerged as the only significant confounding factor, demonstrating a notable influence on the model\u0026rsquo;s performance in predicting patient outcomes for those who received immunotherapy (P\u0026thinsp;=\u0026thinsp;0.047) (Supplementary Table\u0026nbsp;1). This finding underscores the importance of considering disease stage as a critical factor when assessing the prognostic value of biomarkers in immunotherapy-treated patients.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eBlood proteomic biomarkers offer a real-time and non-invasive way to capture dynamic tumour and immune-related signals to better guide disease stratification and therapeutic outcomes for surgery and targeted therapies. We profiled the plasma proteome of 56 NSCLC patients at 87 timepoints, covering various clinical modalities including before and after lung resection surgery, recurrent versus non-recurrent post-surgery, and pre- versus post-immunotherapy. Our analysis used two plasma proteomics platforms, NULISAseq (Alamar Biosciences) and SomaScan (SomaLogic). By applying false discovery-corrected feature-selection methods, we identified distinct protein signatures associated with different clinical groups, including surgical, recurrence, and ICI therapy.\u003c/p\u003e\u003cp\u003eNULISAseq employs a highly sensitive dual-antibody sandwich approach with DNA barcoding and NGS, achieving attomolar sensitivity by significantly reducing background signal. Being highly sensitive enables this platform to detect low-abundance proteins like cytokines and immune regulators using a smaller panel of 250 probes. In contrast, SomaScan uses aptamer technology, where numerous oligonucleotide probes cover 7596 proteins with high throughput. These differences can lead to variability in protein detection between the two platforms, where aptamer and antibody-based assays may indicate variable correlations depending on target characteristics [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. SomaScan's broader coverage facilitated the detection of a diverse array of proteins, including those involved in extracellular matrix remodeling and metabolic regulation. Our direct comparison of these platforms highlights their complementary strengths and specific instances where their detection capabilities vary. The notable cross-platform correlation emphasizes the robustness and reproducibility of our findings, pointing to that both platforms effectively capture comparable biological variations, despite differences in technology, scale of measurements, and panel focus. In overlapping regions of the two protein panels, correlation of essential immune, vascular, and tissue remodeling markers were found across the two independent proteomic platforms. These include chemokines that facilitate immune cell recruitment (CCL28, CCL17, CXCL2, CCL5), as well as mediators involved in extracellular matrix remodeling (MMP1) and angiogenesis (PDGFA, PDGFB, ANGPT1). These variations have noticeable implications for biomarker discovery and interpretation, emphasising the importance of understanding platform-specific characteristics when selecting biomarkers for translational studies.\u003c/p\u003e\u003cp\u003eDetectable biomarkers in the blood, such as those involved in inflammation, coagulation factors, and stress hormones, varied between pre-surgery and post-surgery, provide insights into the biological processes associated with surgical intervention. We observed a reduction in IL-36G in post-surgery samples, a pro-inflammatory cytokine known for activating CD8\u003csup\u003e+\u003c/sup\u003e T cells and NK cells, which may reflect a dampening of systemic inflammation following tumour removal [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. Conversely, we found elevated circulating levels of MUC16 after surgery, a glycoprotein expressed on the surface of various epithelial cells, reported to be higher in different types of lung cancer, particularly in advanced stages [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. We observed, post-surgery downregulation of CLEC3B, KXD1 and TGFBI alongside upregulation of SHC1 and THBS3 [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e], with pathway analysis results further supporting the critical role of cytokines and extracellular matrix components in distinguishing pre-surgery from post-surgery conditions. Reduction of CLEC3B and TGFBI, both associated with tumour ECM remodelling and tissue homeostasis), together with reduced KXD1, linked to lysosomal function, may reflect early systemic responses to surgical intervention. In parallel, increased SHC1 and THBS3, which are involved in wound healing, angiogenesis, and ECM remodelling, likely indicate active tissue repair and remodelling processes [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. These expression changes are consistent with biological adaptations following surgery and may reflect wound healing and tissue reconstruction dynamics.\u003c/p\u003e\u003cp\u003eNSCLC recurrence rate following surgery remains a clinical challenge, with rates ranging from 15\u0026ndash;20% [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. Improving postoperative monitoring and treatment strategies depends on the identification of recurrence-associated biomarkers [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. We found biomarkers involved in immune regulation and inflammation (FCER2, FCRLA and SLITRK2), metabolic adaptation (COX7A2L), cell signalling and growth regulation (FGF19), and extracellular matrix regulation (SPOCK2) associated with tumour recurrence. While no study to date has specifically explored the plasma levels of FCER2, FCRLA and SLITRK2 in lung cancer following surgery, their higher expression within tumour tissue has been linked to better overall survival, reduced risk of metastasis, and more favourable outcomes. In contrast, we found the opposite trend, with patients who experienced recurrence after surgery having lower levels of these markers compared to those who remained cancer-free. Since COX7A2L participates in the stabilisation of mitochondrial respiratory supercomplexes, it is expected to decrease in the hypoxic environment of the TME, as reported by Hollinshead et al., who observed reduced mRNA and protein levels in lung cancer cells [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. Although the role of circulating COX7A2L in post-surgery lung cancer recurrence has not been specifically investigated, we observed higher plasma levels of this marker in patients who experienced tumour recurrence compared to those who remained disease-free. A similar opposite trend between previous tissue studies and our plasma profiling in recurrent cases was also seen for SPOCK2 [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. In contrast, and in line with our plasma findings, Chen et al. reported elevated tissue expression of FGF19 as an independent prognostic indicator for recurrence following NSCLC surgical resection [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eImmunotherapy has transformed lung cancer care, yet predictors of treatment response remain intangible. We profiled baseline plasma samples from patients prior to ICI therapy and categorised biomarkers according to RECIST-defined responses. Among the markers that were elevated in non-responders (SD/PD) relative to responders (CR/PR), there is supporting prior publications linking IL-6 [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e], IL-2RA [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e], SPP1 [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e], CD200R1 and CCL19 to systemic inflammation, immune suppression, and/or poor outcomes.\u003c/p\u003e\u003cp\u003eIL-6 is reported to have contrasting effects on immunotherapy. In one cohort of 125 advanced NSCLC patients, low baseline IL-6 (\u0026lt;\u0026thinsp;13.1 pg/mL) was associated with significantly higher objective response and disease control rates and longer PFS/OS under anti-PD-1/PD-L1 therapy [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e], in line with our findings. On the other hand, another study found that patients with high IL-6 had poor outcomes (median PFS\u0026thinsp;~\u0026thinsp;1.9 vs 6.3 months) [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. Our finding that circulating IL-2RA is elevated in non-responders confirms the results of a previous study that found high baseline IL-2RA serum levels predicted poor PD-1/PD-L1 therapy response [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. SSP1 is a pleiotropic secreted protein overexpressed in lung cancer, which promotes tumour growth, metastasis and creates an immunosuppressive microenvironment [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]. We found that higher baseline SPP1 levels were linked to worse ICI response, which is supported by other studies showing high pre-treatment serum SPP1 correlates with poor clinical response and higher mortality in NSCLC patients treated with nivolumab and pembrolizumab [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eIn addition to these known markers, we identified CD200R1 and CCL19 as unique immunotherapy response markers not previously reported in plasma studies. CD200R1 is a receptor on myeloid/immune cells that transmits inhibitory signals from the CD200 ligand. CCL19 attracts CCR7\u003csup\u003e+\u003c/sup\u003e T cells and dendritic cells, with higher expression in the TME correlating with enhanced immune infiltration and better clinical outcomes [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]. Their opposing immunological roles point to the complex regulatory balance and underexplored pathways within the TME that shape the response to therapy and influence outcomes.\u003c/p\u003e\u003cp\u003eTo extend our findings beyond baseline predictors, we examined circulating markers in 11 patients with matched pre- and post-ICI therapy plasma samples. We identified six markers: CCL21, CD276, IL-2RA, IL-24, IL-1RL1 and AGER, with only IL-2RA having been previously linked to treatment outcome. CCL21, a chemokine involved in lymphocyte trafficking and dendritic cell homing, has not been reported in plasma-based lung cancer studies, though higher plasma levels have been observed in melanoma patients who responded to immunotherapy. CD276, considered an immune checkpoint molecule, was differentially expressed before and after treatment. Torres-Mart\u0026iacute;nez \u003cem\u003eet al\u003c/em\u003e. reported significantly higher levels of soluble CD276 in baseline samples of non-responding NSCLC patients by ELISA [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]. IL-24, which has immune-modulatory and tumour-suppressive characteristics, was also differentially expressed, with responders in our cohort showing higher baseline IL-24 compared to non-responders. To our knowledge, no study has compared plasma IL-24 levels before and after immunotherapy. However, Zhang \u003cem\u003eet al\u003c/em\u003e. found that IL-24 improved prognosis in NSCLC patients, with serum IL-24 levels negatively correlating with TNM classification, consistent with a tumour-suppressive role in NSCLC [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. The plasma level of IL-1RL1 was found to have increased at the time of relapse in NSCLC patients with cancer-associated cachexia (CAC) compared with the non-CAC group, showing a negative correlation of this marker within the circulation with the tumour presence [\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e]. Its reduction in the follow-up samples upon immunotherapy in our cohort is also proving the same trend for this protein.\u003c/p\u003e\u003cp\u003eFinally, we assessed AGER, encoding the Receptor for Advanced Glycation End-products (RAGE), a multiligand pattern-recognition receptor abundantly expressed in lung alveolar cells and involved in inflammation and tissue stress responses. Multiple studies have reported that soluble RAGE negatively correlates with lung cancer, with serum levels decreasing during disease progression [\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e]. Although we observed a similar trend, none of these studies investigated the impact of immunotherapy on circulating RAGE levels.\u003c/p\u003e\u003cp\u003eExisting predictors of response, such as PD-L1 expression, are inconsistent [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. By employing a Cox proportional-hazards model on pretreatment plasma proteomes, alongside the application of the Stabl framework to reduce false discoveries from high-plex panels, we identified a four-protein signature for overall survival: CEACAM5, PTX3, FGF23, and AREG. Notably, higher baseline levels of all these proteins predicted worse outcomes in our cohort. CEACAM5, a member of the carcinoembryonic antigen (CEA) family, acts as an oncofetal growth factor and cell-adhesion molecule. Multiple studies have connected higher preoperative or pretreatment CEA levels with worse prognosis in NSCLC [\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e]. Dall'Olio \u003cem\u003eet al\u003c/em\u003e. further showed that a post-treatment drop in CEA was associated with longer survival under PD-1/PD-L1 therapy [\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e]. Although some reports differ on how circulating CEA relates to therapy response [\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e], most available evidence aligns with our findings, showing that higher baseline CEA correlates with greater risk of poor survival [\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e].\u003c/p\u003e\u003cp\u003ePTX3 (Pentraxin-3) is involved in inflammation and innate immunity, released by myeloid and stromal cells in response to inflammatory cytokines, where it modulates angiogenesis, complement activation, and tissue remodeling. Elevated PTX3 levels, in both tumour tissue and circulation, predict worse overall survival across several malignancies [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e]. While PTX3 has not been extensively studied in lung-cancer immunotherapy, Diamandis \u003cem\u003eet al\u003c/em\u003e. recently found that plasma PTX3 may differentiate lung cancer patients from high-risk smokers with performance comparable to current lung markers [\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e]. This aligns with our observation that elevated baseline PTX3 predicts poor survival, possibly reflecting a pro-tumour inflammatory environment. FGF23 has not been reported in prior NSCLC plasma proteome studies, nor has its association with survival upon immunotherapy been explored. Although not considered a \"classic\" lung cancer marker, higher circulating FGF23 has been linked to shorter overall survival in patients with cancers involving bone metastases [\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e]. Finally, AREG, an EGF-family ligand and autocrine growth factor that binds EGFR to drive epithelial proliferation, was selected by our model, with overexpression correlating with shorter overall survival in NSCLC patients. AREG is secreted by lung tumour-infiltrating dendritic cells in the tumour microenvironment, promoting cancer cell growth and metastasis [\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eFrom pre- and post-surgery comparisons to recurrence risk, immunotherapy response, and survival outcomes, our plasma proteomics analysis identified significant biological patterns across the examined clinical modalities, particularly immune-related markers. IL-6, IL-2RA, LAG3, CD276, CCL19, and CCL21 were among the immune-regulatory proteins most frequently selected across models, indicating a broad role of systemic immune regulation in both treatment response and disease progression. In various instances, poorer outcomes were associated with markers such as SPP1 and AREG, which are known to influence immune suppression and tumour\u0026ndash;stromal interactions. Strong predictors of survival also included metabolic or checkpoint-related markers like FGF23 and CEACAM5, along with inflammatory mediators such as PTX3. In comparisons between surgery and recurrence, tissue remodeling and extracellular matrix-associated proteins, including THBS3 and CLEC3B, were more prominent, suggesting a potential role for them in micrometastatic activity or tumour resection response. Furthermore, all markers were selected using two statistical models, adaptive Lasso and Stabl, which enhances the likelihood that they are relevant despite the large feature space induced by these high-plex panels. We demonstrated how employing two independent high-dimensional technologies can uncover complementary and therapeutically significant plasma signals by combining Alamar's NULISAseq platform's ultra-sensitive quantification of low-abundance proteins with SomaScan's extensive proteome coverage. In addition to emphasising immunological and stromal pathways that are common across lung cancer stages, these results collectively provide a foundation for developing reliable blood-based biomarker panels for treatment stratification and personalised therapy.\u003c/p\u003e\u003cp\u003eThis study has some limitations to take into consideration. The small sample size may limit the generalizability of these results, particularly when considering recurrence or treatment response groups. Although we utilised two complimentary proteomics technologies, each has advantages and disadvantages, and variations in target coverage or sensitivity may affect the outcomes, especially when comparing platforms. For instance, the NULISA panel only partially overlaps with the more general SomaScan panel since it mainly focuses on proteins linked to inflammation. Additionally, we studied plasma samples from only two timepoints, before and after surgery and immunotherapy, so we can only get an overview of the changes in protein levels over time. Furthermore, we are unable to directly connect the changes in plasma proteins to the processes occurring within the \u003cem\u003etumour\u003c/em\u003e or its microenvironment as we did not incorporate corresponding tissue analyses. Lastly, even with the use of well-established normalization techniques, we may not have been able to completely control for variables like sample processing, systemic inflammation, or other medical conditions that could affect plasma protein levels.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn summary, our study has identified a number of notable biomarkers using a plasma proteomics screen of various NSCLC clinical groups leveraging high-plex and highly-sensitive blood proteomics data and a comprehensive data analysis pipeline.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eConflicts of Interest\u003c/h2\u003e\n\u003cp\u003eJH is a director and GM an employee of SurgeCare, SAS. JM is an employee of Standard Biotools. AK is on the Scientific Advisory Board for Omapix Solutions, Predxbio, Molecular Instruments, and Visiopharm.\u003c/p\u003e\n\u003ch2\u003eCompeting Interests\u003c/h2\u003e\n\u003cp\u003eJH is a director and GM an employee of SurgeCare, SAS. JM is an employee of Standard Biotools. AK is on the Scientific Advisory Board for Omapix Solutions, Predxbio, Molecular Instruments, and Visiopharm.\u003c/p\u003e\n\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\n\u003cp\u003eConcept: MA, MEW, AK\u003c/p\u003e\n\u003cp\u003eExperimentation: VYN, SD, CO, WM, JM1, KOB\u003c/p\u003e\n\u003cp\u003eData analysis: CYN, AK1, CL, JH, GM\u003c/p\u003e\n\u003cp\u003eWriting and critical review: all authors.\u003c/p\u003e\n\u003cp\u003e*AK=Arutha Kulasinghe AK1= Aaron Kilgallon, JM=James Mansfield, JM1 = James Monkman\u003c/p\u003e\n\u003ch2\u003eAcknowledgement\u003c/h2\u003e\n\u003cp\u003eThis study was supported by the Frazer Institute (University of Queensland), Queensland Spatial Biology Centre (Wesley Research Institute), Cure Cancer and The PA Research Foundation. The authors acknowledge the Alamar Biosciences and Somalogic technology access programs.\u003c/p\u003e\n\u003ch2\u003eData Availability\u003c/h2\u003e\n\u003cp\u003eAll relevant experimental data are available from the corresponding authors upon reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eSiegel, R.L., K.D. Miller, and A. Jemal, \u003cem\u003eCancer statistics, 2020\u003c/em\u003e. CA Cancer J Clin, 2020. 70(1): p. 7\u0026ndash;30.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhou, J., et al., \u003cem\u003eGlobal burden of lung cancer in 2022 and projections to 2050: Incidence and mortality estimates from GLOBOCAN\u003c/em\u003e. Cancer Epidemiology, 2024. 93: p. 102693.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCrosby, D., et al., \u003cem\u003eEarly detection of cancer\u003c/em\u003e. Science, 2022. 375(6586): p. eaay9040.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAlbanes, D., et al., \u003cem\u003eThe blood proteome of imminent lung cancer diagnosis\u003c/em\u003e. Nature Communications, 2023. 14(1): p. 3042.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRathod, R., et al., \u003cem\u003eFactors influencing false positive low-dose computed tomography-based lung cancer screening in an urban, safety-net system\u003c/em\u003e. Cancer Research, 2025. 85(8_Supplement_1): p. 7408\u0026ndash;7408.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePatz, E.F., et al., \u003cem\u003eOverdiagnosis in low-dose computed tomography screening for lung cancer\u003c/em\u003e. JAMA internal medicine, 2014. 174(2): p. 269\u0026ndash;274.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMehan, M.R., et al., \u003cem\u003eProtein Signature of Lung Cancer Tissues\u003c/em\u003e. PLOS ONE, 2012. 7(4): p. e35157.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMacklin, A., S. Khan, and T. Kislinger, \u003cem\u003eRecent advances in mass spectrometry based clinical proteomics: applications to cancer research\u003c/em\u003e. Clinical Proteomics, 2020. 17(1): p. 17.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhao, R., et al., \u003cem\u003ePlasma proteomic profiles for early detection and risk stratification of non-small cell lung carcinoma: A prospective cohort study with 52,913 participants\u003c/em\u003e. International Journal of Cancer, 2025.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBar, J., et al., \u003cem\u003eBiological insights from plasma proteomics of non-small cell lung cancer patients treated with immunotherapy\u003c/em\u003e. Frontiers in Immunology, 2024. Volume 15\u0026ndash;2024.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTognetti, M., et al., \u003cem\u003eBiomarker Candidates for Tumors Identified from Deep-Profiled Plasma Stem Predominantly from the Low Abundant Area\u003c/em\u003e. J Proteome Res, 2022. 21(7): p. 1718\u0026ndash;1735.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGeyer, P.E., et al., \u003cem\u003eThe Circulating ProteomeTechnological Developments, Current Challenges, and Future Trends\u003c/em\u003e. Journal of Proteome Research, 2024. 23(12): p. 5279\u0026ndash;5295.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYaghoubi Naei, V., et al., \u003cem\u003eAdvances in novel strategies for isolation, characterization, and analysis of CTCs and ctDNA\u003c/em\u003e. Therapeutic Advances in Medical Oncology, 2023. 15: p. 17588359231192401.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eH\u0026eacute;dou, J., et al., \u003cem\u003eDiscovery of sparse, reliable omic biomarkers with Stabl\u003c/em\u003e. Nature Biotechnology, 2024. 42(10): p. 1581\u0026ndash;1593.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHodi, F.S., et al., \u003cem\u003eEvaluation of Immune-Related Response Criteria and RECIST v1.1 in Patients With Advanced Melanoma Treated With Pembrolizumab\u003c/em\u003e. J Clin Oncol, 2016. 34(13): p. 1510\u0026ndash;7.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMeinshausen, N. and P. B\u0026uuml;hlmann, \u003cem\u003eStability selection\u003c/em\u003e. Journal of the Royal Statistical Society Series B: Statistical Methodology, 2010. 72(4): p. 417\u0026ndash;473.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZou, H., \u003cem\u003eThe adaptive lasso and its oracle properties\u003c/em\u003e. Journal of the American statistical association, 2006. 101(476): p. 1418\u0026ndash;1429.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWu, W., et al., \u003cem\u003eIL-2Rα-biased agonist enhances antitumor immunity by invigorating tumor-infiltrating CD25\u0026thinsp;+\u0026thinsp;CD8\u0026thinsp;+\u0026thinsp;T cells\u003c/em\u003e. Nature Cancer, 2023. 4(9): p. 1309\u0026ndash;1325.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLin, X., et al., \u003cem\u003eRegulatory mechanisms of PD-1/PD-L1 in cancers\u003c/em\u003e. Molecular Cancer, 2024. 23(1): p. 108.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRaffield, L.M., et al., \u003cem\u003eComparison of proteomic assessment methods in multiple cohort studies\u003c/em\u003e. Proteomics, 2020. 20(12): p. 1900278.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBaker, K.J., et al., \u003cem\u003eIL-36 expression is increased in NSCLC with IL-36 stimulation of lung cancer cells promoting a pro-tumorigenic phenotype\u003c/em\u003e. Cytokine, 2023. 165: p. 156170.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKimura, Y., et al., \u003cem\u003eSerum CA125 level is a good prognostic indicator in lung cancer\u003c/em\u003e. British Journal of Cancer, 1990. 62(4): p. 676\u0026ndash;678.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNakamura, H. and T. Nishimura, \u003cem\u003eHistory, molecular features, and clinical importance of conventional serum biomarkers in lung cancer\u003c/em\u003e. Surgery Today, 2017. 47(9): p. 1037\u0026ndash;1059.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLy, D., et al., \u003cem\u003eRole for High-Affinity IgE Receptor in Prognosis of Lung Adenocarcinoma Patients\u003c/em\u003e. Cancer Immunology Research, 2017. 5(9): p. 821\u0026ndash;829.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiang, Y., et al., \u003cem\u003eThe increased expression and aberrant methylation of SHC1 in non\u0026ndash;small cell lung cancer: Integrative analysis of clinical and bioinformatics databases\u003c/em\u003e. Journal of Cellular and Molecular Medicine, 2021. 25(14): p. 7039\u0026ndash;7051.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBertoglio, P., et al., \u003cem\u003eEP07.04-19 Clinical and Molecular Profile and Risk Factors of Stage IA NSCLC Relapsing after Radical Resection\u003c/em\u003e. Journal of Thoracic Oncology, 2023. 18(11): p. S552-S553.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSchuurbiers, M.M.F., et al., \u003cem\u003eRecurrence prediction using circulating tumor DNA in patients with early-stage non-small cell lung cancer after treatment with curative intent: A retrospective validation study\u003c/em\u003e. PLOS Medicine, 2025. 22(4): p. e1004574.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHollinshead, K.E., et al., \u003cem\u003eRespiratory supercomplexes promote mitochondrial efficiency and growth in severely hypoxic pancreatic cancer\u003c/em\u003e. Cell reports, 2020. 33(1).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang, Y., et al., \u003cem\u003eNebulized inhalation of extracellular vesicles containing SPOCK2 suppresses lung adenocarcinoma progression via MAPK inhibition\u003c/em\u003e. Discover Oncology, 2025. 16(1): p. 1\u0026ndash;14.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen, J., et al., \u003cem\u003eEnhanced expression of FGF19 predicts poor prognosis in patients with non-small cell lung cancer\u003c/em\u003e. Journal of thoracic disease, 2021. 13(3): p. 1769.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNaqash, A.R., et al., \u003cem\u003eIncreased interleukin-6/C-reactive protein levels are associated with the upregulation of the adenosine pathway and serve as potential markers of therapeutic resistance to immune checkpoint inhibitor-based therapies in non-small cell lung cancer\u003c/em\u003e. J Immunother Cancer, 2023. 11(10).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSiemiątkowska, A., et al., \u003cem\u003eSerum sCD25 Protein as a Predictor of Lack of Long-Term Benefits from Immunotherapy in Non-Small Cell Lung Cancer: A Pilot Study.\u003c/em\u003e Cancers (Basel), 2021. 13(15).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRamoni, D., et al., \u003cem\u003eCirculating Osteopontin Predicts Clinical and Radiological Response in First-Line Treatment of Advanced Non-Small Cell Lung Cancer\u003c/em\u003e. Lung, 2024. 202(2): p. 197\u0026ndash;210.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKang, D.H., et al., \u003cem\u003eBaseline Serum Interleukin-6 Levels Predict the Response of Patients with Advanced Non-small Cell Lung Cancer to PD-1/PD-L1 Inhibitors\u003c/em\u003e. Immune Netw, 2020. 20(3): p. e27.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSu, S., et al., \u003cem\u003ePredictive value of peripheral blood biomarkers in patients with non-small-cell lung cancer responding to anti-PD-1-based treatment\u003c/em\u003e. Cancer Immunology, Immunotherapy, 2024. 73(1): p. 12.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGu, Q., et al., \u003cem\u003eCCL19: a novel prognostic chemokine modulates the tumor immune microenvironment and outcomes of cancers\u003c/em\u003e. Aging (Albany NY), 2023. 15(21): p. 12369\u0026ndash;12387.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eShao, A. and D.M. Owens, \u003cem\u003eThe immunoregulatory protein CD200 as a potentially lucrative yet elusive target for cancer therapy\u003c/em\u003e. Oncotarget, 2023. 14(1).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTorres-Mart\u0026iacute;nez, S., et al., \u003cem\u003eCirculating Immune Proteins: Improving the Diagnosis and Clinical Outcome in Advanced Non-Small Cell Lung Cancer\u003c/em\u003e. International Journal of Molecular Sciences, 2023. 24(24): p. 17587.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang, K., et al., \u003cem\u003eSerum IL-24 combined with CA125 as screening and prognostic biomarkers for NSCLC\u003c/em\u003e. Cell Biology International, 2024. 48(8): p. 1160\u0026ndash;1168.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAl-Sawaf, O., et al., \u003cem\u003eBody composition and lung cancer-associated cachexia in TRACERx\u003c/em\u003e. Nature Medicine, 2023. 29(4): p. 846\u0026ndash;858.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJing, R., et al., \u003cem\u003eReceptor for advanced glycation end products (RAGE) soluble form (sRAGE): a new biomarker for lung cancer\u003c/em\u003e. Neoplasma, 2010. 57(1): p. 55.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDuan, F., et al., \u003cem\u003eA novel diagnostic model for predicting immune microenvironment subclass based on costimulatory molecules in lung squamous carcinoma\u003c/em\u003e. Frontiers in Genetics, 2022. Volume 13\u0026ndash;2022.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDall'Olio, F.G., et al., \u003cem\u003eCEA and CYFRA 21\u0026thinsp;\u0026ndash;\u0026thinsp;1 as prognostic biomarker and as a tool for treatment monitoring in advanced NSCLC treated with immune checkpoint inhibitors\u003c/em\u003e. Ther Adv Med Oncol, 2020. 12: p. 1758835920952994.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDal Bello, M., et al., \u003cem\u003eThe role of CEA, CYFRA21-1 and NSE in monitoring tumor response to Nivolumab in advanced non-small cell lung cancer (NSCLC) patients\u003c/em\u003e. Journal of Translational Medicine, 2019. 17: p. 1\u0026ndash;10.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMarin-Acevedo, J.A., E.O. Kimbrough, and Y. Lou, \u003cem\u003eNext generation of immune checkpoint inhibitors and beyond\u003c/em\u003e. Journal of Hematology \u0026amp; Oncology, 2021. 14(1): p. 45.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJung, H., et al., \u003cem\u003ePrognostic Value of Pentraxin3 Protein Expression in Human Malignancies: A Systematic Review and Meta-Analysis\u003c/em\u003e. Cancers (Basel), 2024. 16(22).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDiamandis, E.P., et al., \u003cem\u003ePentraxin-3 is a novel biomarker of lung carcinoma\u003c/em\u003e. Clin Cancer Res, 2011. 17(8): p. 2395\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMansinho, A., et al., \u003cem\u003eLevels of Circulating Fibroblast Growth Factor 23 (FGF23) and Prognosis in Cancer Patients with Bone Metastases\u003c/em\u003e. International Journal of Molecular Sciences, 2019. 20(3): p. 695.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHsu, Y.L., et al., \u003cem\u003eLung tumor-associated dendritic cell-derived amphiregulin increased cancer progression\u003c/em\u003e. J Immunol, 2011. 187(4): p. 1733\u0026ndash;44.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"npj-precision-oncology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"npjprecisiononcology","sideBox":"Learn more about [npj Precision Oncology](http://www.nature.com/npjprecisiononcology/)","snPcode":"41698","submissionUrl":"https://submission.springernature.com/new-submission/41698/3","title":"npj Precision Oncology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Plasma proteomics, NSCLC, NULISA, SomaScan, immunotherapy, biomarkers","lastPublishedDoi":"10.21203/rs.3.rs-7118825/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7118825/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eObjectives.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNon-small cell lung cancer (NSCLC) remains a leading cause of cancer-related mortality, with limited biomarkers to guide surgical and immunotherapeutic intervention. This study aimed to identify plasma proteomic signatures associated with surgical resection, recurrence, immunotherapy response, and survival by leveraging both high-plex and high-sensitivity proteomics technologies.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTwo complementary plasma proteomics platforms, SomaScan (quantifying 7596 proteins) and NULISAseq (quantifying 250 inflammation-related proteins), were used to profile the blood at 87 timepoints from 56 NSCLC patients. Samples were collected longitudinally: pre- and post-surgery (n = 20), pre- and post-immune checkpoint inhibitor (ICI) therapy (n = 11), and at baseline prior to ICI (n = 25). Proteomics data were analysed using adaptive Lasso regression (ALasso) and the Stabl algorithm for robust selection of a minimal number of differentially expressed proteins that collectively modelled the clinical classification or response.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe identified 21 differentially detectable plasma proteins across surgery and ICI treatment. Surgical resection induced measurable changes in plasma proteins, notably higher circulating MUC16 and lower circulating IL36G post-surgery in non-recurrent patients. Recurrent patients had higher plasma levels of COX7A2L, FGF19 and SPOCK2 post-surgery, and lower FCER2, FCRLA and SLITRK2. Patients who responded to ICI therapy had lower levels of baseline IL-6, CCL19, IL-2RA, CD200R1, CRP, LIF, PDCD1, CCL7 and SPP1 prior to ICI therapy, highlighting associations between systemic inflammation and immune regulation. CEACAM5, PTX3, FGF23, and AREG were elevated in patients with worse clinical outcomes and poorer overall survival. Cross-platform comparisons underscored the complementary strengths of SomaScan’s broad coverage and NULISA’s sensitivity in detecting low-abundance, clinically relevant proteins.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusions.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis integrated plasma proteomics study reveals distinct protein signatures associated with NSCLC treatment response and prognosis. These findings support the utility of non-invasive, blood-based proteomic assays for defining biomarker discovery in NSCLC.\u003c/p\u003e","manuscriptTitle":"Dissecting Non-Small Cell Lung Cancer (NSCLC) with Blood Proteomics - From Surgical to Immunotherapeutic Responses","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-01 10:54:27","doi":"10.21203/rs.3.rs-7118825/v1","editorialEvents":[{"type":"communityComments","content":1},{"type":"decision","content":"Revision requested","date":"2026-01-12T04:13:32+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-24T18:15:36+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-07T14:42:35+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"97992002414706115750780476193292436904","date":"2025-10-06T18:18:57+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"103953139870312844620821661651407470823","date":"2025-09-22T23:34:11+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-07-29T08:29:54+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-07-19T15:27:11+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-07-18T02:44:31+00:00","index":"","fulltext":""},{"type":"submitted","content":"npj Precision Oncology","date":"2025-07-14T08:30:15+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"npj-precision-oncology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"npjprecisiononcology","sideBox":"Learn more about [npj Precision Oncology](http://www.nature.com/npjprecisiononcology/)","snPcode":"41698","submissionUrl":"https://submission.springernature.com/new-submission/41698/3","title":"npj Precision Oncology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"270fdb4d-c6b3-4b8a-b71d-5762e24f6372","owner":[],"postedDate":"August 1st, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":52285586,"name":"Health sciences/Biomarkers"},{"id":52285588,"name":"Biological sciences/Cancer"},{"id":52285589,"name":"Biological sciences/Immunology"},{"id":52285590,"name":"Health sciences/Oncology"}],"tags":[],"updatedAt":"2026-04-29T18:38:21+00:00","versionOfRecord":[],"versionCreatedAt":"2025-08-01 10:54:27","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7118825","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7118825","identity":"rs-7118825","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}