Multi-omics dynamic profiling reveals predictive biomarkers for first-line immunochemotherapy in extensive-stage small-cell lung cancer

Liang Zheng; Haoming Xu; Shuyuan Wang; Meili Ma; Fang Hu; Lei Cheng; Jun Lu; Feng Pan; Bo Zhang; Jianlin Xu; Ying Li; Yinchen Shen; Wei Zhang; Runbo Zhong; Tianqing Chu; Baohui Han; Xiaoxuan Zheng; Hua Zhong; Wei Nie; Xueyan Zhang

doi:10.1186/s12967-026-07778-y

. 2026 Feb 7;24:353. doi: 10.1186/s12967-026-07778-y

Multi-omics dynamic profiling reveals predictive biomarkers for first-line immunochemotherapy in extensive-stage small-cell lung cancer

Liang Zheng ^1,^#, Haoming Xu ^1,^#, Shuyuan Wang ¹, Meili Ma ¹, Fang Hu ^3,⁴, Lei Cheng ¹, Jun Lu ¹, Feng Pan ¹, Bo Zhang ¹, Jianlin Xu ¹, Ying Li ¹, Yinchen Shen ¹, Wei Zhang ¹, Runbo Zhong ¹, Tianqing Chu ¹, Baohui Han ^1,^✉, Xiaoxuan Zheng ^1,^2,^5,^✉, Hua Zhong ^1,^✉, Wei Nie ^1,^✉, Xueyan Zhang ^1,^✉

PMCID: PMC12977650 PMID: 41654892

Abstract

Background

Extensive-stage small-cell lung cancer (ES-SCLC) is associated with a poor prognosis. Although first-line immunochemotherapy improves clinical outcomes, robust prognostic biomarkers for this treatment modality remain unavailable. The aim of this study was to identify non-invasive, easily accessible, and dynamically monitored biomarkers of ES-SCLC by machine learning integrating serum metabolomics, lipidomics, and proteomics at multiple time points.

Methods

A total of 816 serum samples were collected from ES-SCLC patients receiving first-line immunotherapy combined with chemotherapy or first-line chemotherapy for metabolomics, lipidomics, and proteomics analysis. The immunochemotherapy cohort was randomly divided into training and validation subsets at a 6:4 ratio. Biomarkers were identified using machine learning algorithms, and their prognostic significance was evaluated through receiver operating characteristic (ROC) analysis, Kaplan–Meier survival analysis, and multivariate Cox regression. Potential metabolic pathways and mechanisms were further explored via integrated multi-omic analysis.

Results

The immunochemotherapy exhibited a prolonged median progression-free survival (PFS) and higher objective response rate (ORR) compared to the chemotherapy group. A total of 5 serum metabolites (uric acid, L-aspartate-semialdehyde, dimethisterone, xanthine, L-cysteine), 6 lipids (Cer d18:1/26:0, Cer d18:2/25:0, SM d18:1/20:1, SM d17:1/25:1, DG O-18:1_16:0, PS 18:0_24:0), and 3 proteins (ACIN1, ACSL4, PHGDH) were identified and constructed into independent prognostic models. Among patients receiving immunochemotherapy, those categorized as low-risk based on the model demonstrated significantly longer PFS compared with those in the high-risk group. These prognostic signatures also retained predictive value in patients who underwent second-line treatment with anlotinib plus immunochemotherapy. Integrated analysis revealed that glycine, serine, and threonine metabolism was the commonly enriched pathway across all three omics layers. Notably, PHGDH (protein), L-aspartate-semialdehyde and L-cysteine (metabolites), and PS (18:0_24:0) (lipid), key elements in this pathway, were all incorporated in the predictive model. In addition, models of the composition of these substances after one cycle of treatment can still predict the prognosis of patients.

Conclusion

In this study, we constructed and validated a set of non-invasive, dynamically monitorable prognostic models (containing 5 metabolites, 6 lipids, and 3 proteins) using machine learning by integrating multiple time point data from the serum metabolome, lipid panel, and proteome to accurately distinguish the prognostic risk of patients with ES-SCLC receiving immunochemotherapy. PFS was significantly prolonged in patients in the low-risk group, and this model remains predictive in the subsequent second-line treatment with anlotinib in combination with immunochemotherapy. Glycine-serine-threonine metabolic pathway may be the key mechanism, of which PHGDH, L-aspartate semialdehyde, L-cysteine and PS (18:0_24:0) are the core predictors. This study provides the first multi-omics dynamic prognostic tool for ES-SCLC immunochemotherapy and reveals potential therapeutic targets.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-026-07778-y.

Keywords: Small-cell lung cancer, Biomarkers, Metabolomics, Lipidomics, Proteomics

Introduction

Small-cell lung cancer (SCLC) accounts for approximately 10%–15% of all lung cancers and represents a major cause of cancer-related mortality worldwide [1]. It is characterized by rapid progression, early metastasis, and poor prognosis. Clinically, SCLC is classified into limited-stage and extensive-stage disease [2], with the majority of patients diagnosed at the extensive stage. Chemotherapy has traditionally been the cornerstone of treatment for extensive-stage SCLC (ES-SCLC) [3]. Specifically, platinum-based regimens—etoposide combined with cisplatin or carboplatin—have constituted the standard of care for many years [4]. However, these regimens are associated with suboptimal therapeutic outcomes, limited antitumor efficacy, rapid disease progression, high recurrence rates, and a dismal five-year survival rate [5].

In recent years, the treatment paradigm for SCLC has shifted from chemotherapy alone toward immunochemotherapy, which has shown considerable promise in improving clinical outcomes [6]. Immune checkpoint inhibitors (ICIs), particularly those targeting PD-1/PD-L1 such as atezolizumab, durvalumab, and serplulimab, in combination with chemotherapy, have become standard first-line treatments for ES-SCLC [7]. Although this combination significantly improves overall survival (OS) and sets a new benchmark for first-line therapy, only a subset of patients derive long-term benefit [8]. Therefore, the identification of predictive biomarkers to determine which patients are likely to respond favorably to immunochemotherapy is crucial for advancing SCLC treatment.

Programmed death-ligand 1 (PD-L1) expression is a well-established predictive biomarker for ICI efficacy in several cancers [9]. However, in SCLC, PD-L1 expression is typically low on tumor cells, limiting its predictive utility [10]. While some studies have suggested that high PD-L1 expression may correlate with better responses to immunotherapy, its prognostic value in SCLC remains controversial [11, 12]. Similarly, although tumor mutation burden (TMB) is considered a potential predictive biomarker, its relevance in SCLC is not well-established [13]. For instance, the CASPIAN and IMpower133 trials reported no significant association between TMB and immunotherapy efficacy in SCLC [14, 15]. Research on blood-based TMB (bTMB) is also limited, and there is insufficient evidence to support its use as an independent predictive biomarker in SCLC [16]. Consequently, despite their predictive roles in non-small-cell lung cancer (NSCLC), the clinical application of PD-L1 and TMB in SCLC remains challenging. Other potential biomarkers include immune cell markers such as CD8A, CD4, and MHC class I and II [17], as well as circulating tumor DNA (ctDNA) and circulating tumor cells (CTCs), which may reflect treatment response [18, 19]. However, studies evaluating these biomarkers specifically in SCLC are sparse, and their predictive value for immunotherapy efficacy remains to be conclusively demonstrated.

In recent years, multi-omic technologies have significantly advanced biomarker research for immunotherapy in SCLC. Some investigations have suggested that specific genetic or protein alterations may predict immunotherapy response [20, 21]. Nevertheless, alterations in a single gene or protein are insufficient to fully elucidate mechanisms of resistance to immunotherapy in SCLC. Comprehensive analyses integrating proteomics, genomics, and transcriptomics allow for deeper insights into the molecular features and immune microenvironment of SCLC [22, 23], providing a theoretical foundation for precision immunotherapy. Additionally, limited accessibility to tumor tissue and the high heterogeneity of SCLC hinder the clinical application of tissue-based biomarkers. In contrast, liquid biopsy offers several advantages, including non-invasiveness, reproducibility, and the ability for dynamic monitoring. Therefore, it is essential to explore prognostic biomarkers for ES-SCLC treated with immunochemotherapy through the detection of lipids, metabolites, and proteins in serum samples.

In addition, although ES-SCLC initially responds well to treatment, the majority of patients experience relapse within a short period, accompanied by reduced drug sensitivity and poor post-relapse prognosis. Previous studies have shown that anlotinib is a novel multi-target tyrosine kinase inhibitor. It suppresses tumor angiogenesis and cell proliferation by blocking vascular endothelial growth factor receptor (VEGFR), fibroblast growth factor receptor (FGFR), platelet-derived growth factor receptor (PDGFR), and stem cell factor receptor (c-kit) [24, 25]. In real-world clinical practice, we observed that combining anlotinib with PD-1/PD-L1 inhibitors and chemotherapy as second-line therapy for ES-SCLC resulted in prolonged median progression-free survival (PFS) with manageable safety profiles [26]. Consequently, we are interested in elucidating the underlying synergistic mechanisms of anlotinib in combination with immunotherapy. The promising efficacy and safety of this combination therapy in second-line treatment suggest that anlotinib may offer additional therapeutic options for patients with ES-SCLC.

To identify non-invasive, dynamically monitorable biomarkers for first-line immunochemotherapy in ES-SCLC and to explore potential mechanistic links at the proteomic, metabolomic, and lipidomic levels, we dynamically investigated hematological changes at multiple time points using serum-based untargeted metabolomics, broadly targeted lipidomics, and ultrafast proteomics. Our study identified multiple metabolites, lipids, proteins, and metabolic pathways as dynamic prognostic indicators for first-line immunochemotherapy in ES-SCLC. Using machine learning approaches, we established biomarker panels across different omics layers to stratify patients by prognosis, including those receiving first-line immunotherapy plus chemotherapy and those undergoing second-line treatment with anlotinib. Finally, through comprehensive multi-omics integration, we elucidated altered metabolic pathways and proposed potential therapeutic strategies aimed at improving the prognosis and overcoming resistance to immunochemotherapy.

Methods

Patient selection

In this study, between January 2021 and September 2023, we collected a total of 8,675 patients who were diagnosed with lung cancer by histological pathology and received treatment at Shanghai Chest Hospital. These patients were then screened in accordance with the inclusion and exclusion criteria, and the follow-up period ended in July 2024. The screening process was shown in Supplementary Fig. 1, and finally we included a longitudinal cohort of 204 ES-SCLC patients and tested a total of 816 samples at different time points. Of these, 144 patients received first-line immunotherapy combined with chemotherapy, while 60 patients received chemotherapy alone. The study was conducted in accordance with all relevant ethical guidelines and was approved by the Institutional Review Board and Ethics Committee of Shanghai Chest Hospital (Reference Number: LS1808). Informed consent was obtained from all participants.

Inclusion criteria were as follows: (1) histologically or cytologically confirmed diagnosis of small-cell lung cancer; (2) diagnosis of ES-SCLC based on the 9th edition of the TNM staging system, with surgical intervention not recommended; (3) presence of measurable disease; (4) receipt of first-line immunotherapy plus chemotherapy or chemotherapy alone; and (5) an Eastern Cooperative Oncology Group (ECOG) performance status (PS) score of 0–1. Exclusion criteria included: (1) failure to complete necessary systemic evaluations; (2) prior pulmonary surgery; (3) presence of severe active infection; (4) incomplete clinical or follow-up data; (5) presence of other primary active malignancies; and (6) history of autoimmune diseases, severe cardiopulmonary dysfunction, or other serious comorbidities.

Patient treatment and specimen collection

Patients receiving first-line immunochemotherapy were treated with durvalumab, atezolizumab, adebrelimab, or serplulimab, in combination with etoposide and cisplatin or carboplatin. Patients receiving first-line chemotherapy alone were administered etoposide combined with either cisplatin or carboplatin. Treatment cycles were conducted every 3–4 weeks, with individualized dosing based on body surface area and patient tolerance. Treatment continued until the occurrence of intolerable adverse reactions, disease progression, or death. Disease progression was assessed prior to each treatment cycle by two experienced physicians using imaging modalities such as non-contrast chest CT, abdominal ultrasound, contrast-enhanced brain MRI, or bone scans.

Additionally, comprehensive clinical data were collected, including age, sex, smoking history, ECOG PS, T stage, N stage, metastasis status, history of radiotherapy, adverse events, PD-L1 expression, height, weight, and medical history (coronary artery disease, diabetes, and hypertension). Peripheral venous blood samples were collected at two time points, with baseline within 3 days before the start of treatment and another time point after one course of treatment, and the date of collection was recorded. Serum samples were stored at -80 °C in the hospital biobank. Eligible samples were retrieved for metabolomic, lipidomic, and proteomic analyses.

Untargeted metabolomics

Serum samples were initially removed from the − 80 °C freezer, thawed on ice, and vortex-mixed. Fifty microliters of thawed serum was transferred into a centrifuge tube, followed by the addition of 300 µL of 20% acetonitrile-methanol internal standard extraction solution. Samples were vortexed for 3 min and centrifuged at 12,000 rpm for 10 min at 4 °C. Subsequently, 200 µL of the supernatant was transferred to a new centrifuge tube, incubated at − 20 °C for 30 min, and centrifuged again for 3 min. A final volume of 180 µL of supernatant was transferred into the liner tube of the injection vial for LC-MS/MS analysis.

Metabolite separation and detection were performed using liquid chromatography–tandem mass spectrometry (LC-MS/MS) on a Waters ACQUITY Premier HSS T3 column (1.8 μm, 2.1 mm × 100 mm). Mobile phase A consisted of 0.1% formic acid in water, and mobile phase B was 0.1% formic acid in acetonitrile. The column temperature was maintained at 40 °C, with a flow rate of 0.4 mL/min and an injection volume of 4 µL. Mass spectrometry was conducted in both positive and negative electrospray ionization (ESI) modes, with ionization voltages of 3500 V and 3200 V, respectively. The scan range was m/z 75–1000, with a resolution of 35,000. This method ensured accurate qualitative and quantitative metabolite analysis. Quality control (QC) samples were inserted after every 10 test samples to monitor instrument stability and correct for potential signal variation over time. These QC samples were used both as internal controls to assess analytical performance and as references for normalization of raw metabolomics data.

Broadly targeted lipidomics

Serum samples were removed from the − 80 °C freezer and thawed on ice until fully liquefied. Each sample was vortexed for 10 s to ensure homogeneity. Fifty microliters of the thawed sample was transferred into a centrifuge tube, followed by the addition of 1 mL of an extraction solution containing internal standard lipids (methyl tert-butyl ether: methanol = 3:1, v/v), and the mixture was vortexed for 15 min. Subsequently, 200 µL of water was added and vortexed for 1 min. The mixture was centrifuged at 12,000 rpm for 10 min at 4 °C. After centrifugation, 200 µL of the upper phase was transferred to a new centrifuge tube and evaporated to dryness. The residue was then reconstituted in 200 µL of lipid reconstitution solution (acetonitrile: isopropanol = 1:1, v/v), vortexed for 3 min, and centrifuged at 12,000 rpm for 3 min. The resulting supernatant was subjected to LC-MS/MS analysis.

The analytical instrumentation consisted of Ultra-Performance Liquid Chromatography (UPLC) coupled with tandem mass spectrometry (MS/MS). Mobile phase A was 0.1% formic acid in water, and mobile phase B was 0.1% formic acid in acetonitrile. Gradient elution was employed to optimize separation conditions for efficient lipid separation. The mass spectrometry conditions included ESI at 500 °C, a voltage of + 5500 V in positive ion mode and − 4500 V in negative ion mode. Ion source gas settings were as follows: gas 1 (GS1), 45 psi; gas 2 (GS2), 55 psi; curtain gas (CUR), 35 psi. Collision-induced dissociation (CAD) parameters were set to medium. Each ion pair was scanned and detected in the triple quadrupole system according to the optimized declustering potential (DP) and collision energy (CE).

Lipid identification was based on the self-constructed MetWare Database (MWDB), utilizing retention time (RT) and parent–daughter ion information for qualitative analysis. Quantification was performed using multiple reaction monitoring (MRM) mode on the triple quadrupole mass spectrometer. In MRM mode, precursor (parent) ions of target compounds were selectively filtered, eliminating interference from other molecular weights. After collision-induced fragmentation, characteristic fragment ions were selectively filtered to eliminate non-target interference, thereby improving accuracy and reproducibility of quantification. Following MS analysis, peak areas of the lipid species were integrated, and the corresponding peaks across different samples were corrected and normalized for further analysis.

Ultra-fast hematoproteomics

To overcome the limitations imposed by high-abundance proteins in the blood proteome, numerous depletion technologies have been developed. Among them, the nano-graphene bead-based enrichment method for low-abundance proteins has demonstrated superior performance due to its species-independent applicability and high throughput. The method involves the use of functional biomagnetic beads to selectively adsorb low-abundance proteins, forming protein coronas encapsulating nanoparticles, followed by protein detection. In this study, ultra-fast hematoproteomics was performed using a protocol based on functional biomagnetic bead enrichment and Data-Independent Acquisition (DIA) on an Orbitrap™ Astral™ mass spectrometer.

Initially, samples were carefully retrieved and thawed on ice to preserve protein integrity. Phenylmethylsulfonyl fluoride (PMSF) was added to a final concentration of 1 mM, followed by vortex mixing to ensure uniform dispersion. Low-abundance blood proteins were then enriched using the EasyPept™ nanomagnetic bead kit. Reductive alkylation was subsequently conducted directly on the magnetic beads. Proteins were enzymatically digested into peptides using trypsin. The resulting peptides were desalted using a C18 column (Millipore, Billerica, MA), and peptide concentration was determined using a BCA protein assay kit.

Peptide separation was carried out on the Vanquish Neo UHPLC system (Thermo Fisher Scientific). Mobile phase A consisted of 0.1% formic acid in water, and mobile phase B was acetonitrile with 0.1% formic acid. A trap-and-elute dual-column setup was employed, using a PepMap Neo Trap Cartridge (300 × 5 mm, 5 μm) as the trapping column and an Easy-Spray™ PepMap™ Neo UHPLC column (150 × 15 cm, 2 μm) as the analytical column. The column temperature was maintained at 55 °C, and the injection volume was set to 200 ng. The flow rate was 2.5 µL/min, with an effective gradient elution time of 6.9 min and a total run time of 8 min. After separation via nano-flow HPLC, peptide analysis was performed using the Orbitrap Astral high-resolution mass spectrometer.

Raw mass spectrometry data were processed using DIA-NN (v1.8.1) in a library-free mode. A FASTA database containing 82,493 protein sequences was used to generate a spectral library via deep learning-based neural network algorithms. The Match Between Runs (MBR) function was applied to construct a spectral library from the DIA data, which was then used for reanalysis. The false discovery rate (FDR) was controlled at less than 1% at both the protein and precursor ion levels. Only identifications meeting this threshold were retained for further quantification.

Study design

We initially compared survival outcomes between the immunotherapy plus chemotherapy group and the chemotherapy-only group. The primary endpoint of the study was PFS, and the secondary endpoint was overall response rate (ORR). PFS was defined as the time from the first day of treatment to the date of disease progression or death. ORR was defined as the proportion of patients who achieved a complete response (CR) or partial response (PR) among all treated individuals. Serum samples collected from all patients prior to treatment were subjected to untargeted metabolomics and lipidomics analyses to preliminarily investigate differences in metabolic profiles—specifically lipid and non-lipid metabolites—between the two treatment groups.

We subsequently determined the optimal cutoff value for PFS in the immunotherapy plus chemotherapy group by plotting the sensitivity against (1-specificity) using the receiver operating characteristic (ROC) curve. The sum of sensitivity and specificity minus 1 (i.e., Youden index) was calculated using whether a patient ‘s PFS progressed to a binary outcome, ie, state variable, PFS time as a test variable, and the point with the largest Youden index was selected as the cutoff value. With a sensitivity of 95.2% and specificity of 56.1% at 7.5 months PFS, the resulting Youden index of 0.513, which is the maximum Youden index, PFS of 7.5 months was determined to be the optimal threshold. Based on this threshold, patients with PFS ≥ 7.5 months were classified as the response group (R = 74), and those with PFS < 7.5 months as the non-response group (NR = 70), among the 144 patients receiving combination immunotherapy. Metabolomic and lipidomic profiles were compared between these groups, and differential metabolites and lipids were initially identified using the criteria of variable importance in projection (VIP) > 1 and P < 0.05.

Patients were randomly assigned to a training set and a validation set in a 6:4 ratio. Various machine learning methods, including least absolute shrinkage and selection operator (LASSO) regression and random forest (RF), were used to construct predictive models. All data preprocessing and feature selection steps in this study did not use global information at any full-cohort level before data splitting, avoiding the risk of model overfitting due to cross-set information leakage throughout. ROC curves were generated to evaluate model performance. The optimal model, selected based on machine learning performance, was used to calculate risk scores, which were then employed to stratify the overall cohort into high-risk and low-risk groups. Differences in prognosis between these risk groups were compared to further assess model performance. Serum proteomic profiling was performed in second-line patients treated with immunotherapy combined with chemotherapy and anlotinib. A prognostic model based on proteomic data was constructed using the same approach as for metabolomic and lipidomic analyses. The associations among differential proteins, metabolites, and lipids were explored via KEGG pathway co-enrichment and multi-omics correlation analyses.

After comparing baseline patient characteristics, we collected patients’ sera after one cycle of treatment for multi-omics analysis. Patients were also divided into R and NR groups, and volcano plots were used to analyze whether the differential substances contained substances in the previous prediction model, and the relative contents of various substances were counted and compared with the baseline contents to analyze the changes. Finally, patients were divided into low-risk and high-risk groups according to the relative content of these substances after one cycle of treatment, and survival curves were plotted to again verify whether our established prediction model could predict the prognosis of ES-SCLC patients receiving immune combination therapy.

Data analysis and statistical methods

Baseline characteristics were compared using the Chi-square test or Fisher’s exact test, as appropriate. Kaplan-Meier survival curves were generated and compared using the log-rank test. Univariate and multivariate Cox proportional hazards regression models were used to calculate hazard ratios (HRs) and 95% confidence intervals (CIs), with or without adjustment for available prognostic clinical covariates. ROC curves were used to evaluate the predictive power of each variable, with sensitivity plotted against 1-specificity. The area under the curve (AUC) was calculated, with higher values indicating better predictive ability. To estimate the 95% CI of AUC, 1000 bootstrap samples were generated, AUCs were calculated for each, sorted, and the 2.5th and 97.5th percentiles were taken. The results were reported as 95% CI = [CI_lower, CI_upper].

In this study, omics data were log₂-transformed and median-normalized. Several statistical methods were employed to analyze untargeted metabolomics data. Principal component analysis (PCA) was first applied for an overall assessment of metabolic differences across samples. Cluster analysis was then conducted to classify accumulation patterns and assess similarities and differences among samples. Differential metabolite screening combined univariate statistical tests (e.g., t-test P-values) and multivariate analysis (e.g., VIP values from orthogonal partial least squares discriminant analysis [OPLS-DA]) to identify statistically significant features. Functional annotation and pathway enrichment of differential metabolites were performed using the KEGG database to elucidate biological functions and regulatory mechanisms. Analytical approaches for broadly targeted lipidomics and ultra-fast serum proteomics were similar to those used for untargeted metabolomics. Correlations between differential proteins and metabolites were assessed via KEGG pathway enrichment to explore their potential interactions.

Student’s t-test was applied to normally distributed variables with homogeneity of variance between two groups, while the Mann-Whitney U test was used for variables not meeting these assumptions. LASSO regression was conducted using the glmnet, foreign, and tidyr packages in R. Random forest analysis was performed with the varSelRF package. All statistical analyses were carried out using R, GraphPad Prism, or SPSS version 24.0. Figures were prepared using Adobe Illustrator 2022. A two-tailed P-value < 0.05 was considered statistically significant.

Results

Baseline characteristics of patients

The study design and analysis workflow were shown in the Fig. 1. A total of 8675 patients were enrolled in the study, and 204 ES-SCLC patients met our requirements, including 144 patients (70.6%) who received first-line combination therapy and 60 patients (29.4%) who received first-line chemotherapy. Baseline characteristics are summarized in [Supplementary Table 1]. The baseline characteristics were balanced between the immunotherapy plus chemotherapy (Immu + Che) and chemotherapy (Che) groups, as shown in [Supplementary Table 2]. The chemotherapy group was stratified into NR (n = 33) and R (n = 27) subgroups based on the calculated cutoff value, while the Immu + Che group was divided into NR (n = 70) and R (n = 74) subgroups. Baseline characteristics were balanced across all subgroups [Supplementary Table 3]. Furthermore, the Immu + Che group was randomly divided into training (n = 84) and validation (n = 60) sets in a 6:4 ratio using stratified random sampling. The baseline characteristics between NR and R subgroups within both the training and validation sets were also balanced [Supplementary Table 4].

Prognosis of patients receiving different treatments

According to the Kaplan-Meier survival curves in Fig. 2A, the PFS of the Immu + Che group was significantly longer than that of the Che group (P = 0.017). The median PFS was 9.0 months (95% CI, 7.7–10.3) in the Immu + Che group versus 7.0 months (95% CI, 6.1–7.9) in the Che group. In Fig. 2B, after stratification based on the cutoff value, PFS was significantly longer in the R group than in the NR group in both the Immu + Che (13.0 months, 95% CI, 11.4–14.6 vs. 5.0 months, 95% CI, 4.5–5.5; P < 0.001) and Che groups (11.0 months, 95% CI, 10.0–12.0 vs. 5.0 months, 95% CI, 4.1–5.9; P < 0.001). Notably, the R subgroup in the Immu + Che group had a significantly longer PFS than the R subgroup in the Che group (13.0 months, 95% CI, 11.4–14.6 vs. 11.0 months, 95% CI, 10.0–12.0; P = 0.02).

As shown in Fig. 2C, the ORR in the Immu + Che group was 65.3%, significantly higher than that in the Che group (46.7%; P = 0.013). Moreover, the ORR in the Immu + Che-R group was significantly higher than that in the Immu + Che-NR group (81.1% vs. 48.6%; P < 0.001). Additionally, the ORR in the Immu + Che-R group (81.1%) was significantly higher than that in the Che-R group (59.3%; P = 0.025). However, no significant difference in ORR was observed between the Che-NR and Che-R groups (36.4% vs. 59.3%; P = 0.077), nor between the Immu + Che-NR and Che-NR groups (48.6% vs. 36.4%; P = 0.063), as illustrated in Fig. 2D.

To identify clinical factors associated with patient prognosis, we conducted a univariate Cox regression analysis. As shown in Fig. 2E, no clinical characteristics were significantly associated with PFS in the Immu + Che group (P > 0.05). However, a body mass index (BMI) of 18.5–23.9 (HR = 0.236, 95% CI: 0.086–0.649; P = 0.005), BMI > 23.9 (HR = 0.216, 95% CI: 0.077–0.608; P = 0.004), and M stage (HR = 0.432, 95% CI: 0.210–0.887; P = 0.022) were independently associated with PFS (Fig. 2F).

Machine learning approaches to screen differential metabolites

Following metabolomic profiling of serum samples from all patients, we first compared the Immu + Che-NR and Immu + Che-R groups. PCA revealed some differences in distribution between the two groups, although certain regions overlapped, indicating both shared and distinct features (Supplementary Fig. 2A). The OPLS-DA plot demonstrated a relatively dispersed sample distribution and a tendency toward separation, suggesting that the two groups could be differentiated based on their metabolic profiles (Supplementary Fig. 2B).

Differential metabolites between the two groups were identified based on VIP scores > 1 and P values < 0.05, resulting in the initial identification of 81 differential metabolites (Supplementary Fig. 2C). Among these, 54 metabolites were downregulated and 27 were upregulated in the Immu + Che-R group compared with the Immu + Che-NR group. The top 20 metabolites ranked by fold change (FC) are shown in Supplementary Fig. 2D. The most significantly upregulated metabolites were Deethylatrazine, Uric acid, and N-Formyldemecolcine, while the most significantly downregulated metabolites were Glycochenodeoxycholic acid 7-sulfate, Gln-Ile-Tyr-Glu, and Glaucolide A. KEGG pathway enrichment analysis of these differential metabolites indicated potential involvement in glutathione metabolism, as well as glycine, serine, and threonine metabolism (Supplementary Fig. 2E).

We also analyzed differences between the Che-NR and Che-R groups. As shown in the PCA plot (Supplementary Fig. 2F), these groups also exhibited partial overlap with a tendency toward separation, which was more pronounced in the OPLS-DA plot (Supplementary Fig. 2G). Based on the same criteria (VIP > 1, P < 0.05), a total of 99 differential metabolites were identified, with 55 upregulated and 44 downregulated in the Che-R group (Supplementary Fig. 2H). Notably upregulated metabolites included Inosine, Palmitic acid, and Dichlorprop, while Leu-Tyr-Gln-Glu, 4-Oxoglutaramic acid, and 3’-Hydroxyflavanone were among the most significantly downregulated (Supplementary Fig. 2I). These differential metabolites were primarily enriched in metabolic pathways such as fatty acid biosynthesis and the prolactin signaling pathway (Supplementary Fig. 2J).

To refine the set of meaningful differential metabolites identified through initial screening, we randomly divided patients receiving first-line immunotherapy plus chemotherapy into a training set (n = 84) and a validation set (n = 60) using stratified sampling at a 6:4 ratio, ensuring comparable clinical baseline characteristics between the groups. In the training set, we selected differential metabolites specific to the Immu + Che-NR versus Immu + Che-R comparison and excluded those that also showed significant differences in the chemotherapy group, yielding a final list of 25 metabolites (Fig. 3A).

We then applied LASSO regression to this set in the training cohort (Fig. 3B), identifying 11 metabolite combinations that were used to construct a ROC curve. The AUC was 0.942 (95% CI: 0.896–0.988) in the training set and 0.938 (95% CI: 0.873–1.000) in the validation set. In parallel, we used a random forest algorithm to identify an optimal model, resulting in a combination of five metabolites with an AUC of 0.909 (95% CI: 0.849–0.970) in the training set and 0.882 (95% CI: 0.795–0.969) in the validation set, demonstrating strong predictive performance. We also tested a model using the 10 metabolites with the lowest P values from the training set. This model achieved an AUC of 0.898 (95% CI: 0.835–0.962) in the training set and 0.769 (95% CI: 0.646–0.892) in the validation set; however, its performance in the validation set was suboptimal.

For potential clinical applicability, we prioritized models with fewer variables and selected the five-metabolite combination derived from the random forest method. These metabolites were Uric acid, L-Aspartate-semialdehyde, Dimethisterone, Xanthine, and L-Cysteine. ROC curves for the training and validation cohorts are shown in Fig. 3C and D, respectively. Among these, Uric acid and Dimethisterone were significantly downregulated, while L-Aspartate-semialdehyde, Xanthine, and L-Cysteine were significantly upregulated in patients who showed poor response to first-line combination immunotherapy (Fig. 3E). A heatmap of these five differential metabolites confirmed the same expression patterns observed in the violin plot (Fig. 3F).

Machine learning approaches for lipid screening

Similar to the metabolomics analysis, we performed a comparative analysis of broadly targeted lipidomic profiles between the Immu + Che-NR and Immu + Che-R groups. Both PCA (Supplementary Fig. 3A) and OPLS-DA (Supplementary Fig. 3B) plots revealed distinctions in distribution between the two groups, with a clear trend toward separation, although certain regions overlapped, suggesting the presence of both shared and unique lipidomic characteristics.

Based on the criteria of VIP > 1 and P < 0.05, we initially identified 148 differential lipids (Supplementary Fig. 3C). Of these, 145 lipids were upregulated and 3 were downregulated in the Immu + Che-R group compared with the Immu + Che-NR group. The 20 lipids with the most pronounced fold changes are shown in Supplementary Fig. 3D. The most strongly upregulated lipids were Cer (d18:2/24:0 (2OH)), HexCer (d19:1/24:0), and Cer (d16:1/24:0), while the most downregulated lipid was DGDG (14:0/20:5). KEGG pathway enrichment analysis indicated that these differential lipids were mainly enriched in pathways related to sphingolipid metabolism and the sphingolipid signaling pathway (Supplementary Fig. 3E).

In addition, we analyzed lipidomic differences between the Che-NR and Che-R groups. As shown in Supplementary Fig. 3F and Supplementary Fig. 3G, the PCA and OPLS-DA plots revealed both overlapping and distinct distributions between the two groups, indicating differences in lipid profiles associated with chemotherapy efficacy. A total of 59 differential lipids were identified between the Che-NR and Che-R groups based on the screening criteria of VIP > 1 and P < 0.05, among which 3 lipids were upregulated and 56 were downregulated in the Che-R group (Supplementary Fig. 3H). As illustrated in Supplementary Fig. 3I, Cer (t14:2/26:2) was the most significantly upregulated lipid, while FFA (14:1), LPC (16:2), and FFA (12:0) were among the most downregulated. These differential lipids were predominantly involved in metabolic pathways such as sphingolipid metabolism and necroptosis (Supplementary Fig. 3J).

Given the large number of differential lipids identified during preliminary analysis, we employed machine learning approaches to further refine the selection of biologically and clinically relevant lipids. This analysis was conducted within the well-stratified training and validation cohorts. In the training group, 77 differential lipids were found to overlap with those in the Immu + Che-R vs. Immu + Che-NR comparison (Fig. 4A). Among them, 7 lipids also intersected with those differentially expressed between Che-R and Che-NR groups. To eliminate confounding effects of chemotherapy, the remaining 70 lipids were subjected to further screening using machine learning algorithms, including LASSO regression and random forest models.

Using LASSO regression, six candidate lipids were selected: Cer (d18:1/26:0), Cer (d18:2/25:0), SM (d18:1/20:1), Cer (d17:1/25:1), PS (18:0_24:0), and DG (O-18:1_16:0) (Fig. 4B). The AUC for this lipid panel was 0.904 (95% CI: 0.835–0.872) in the training group and 0.874 (95% CI: 0.786–0.962) in the validation group. In contrast, the random forest approach identified two lipids—SM (d18:1/16:1) and SM (d18:1/22:1)—but their predictive performance was suboptimal, with AUCs of 0.728 (95% CI: 0.617–0.838) in the training group and 0.703 (95% CI: 0.570–0.836) in the validation group. Additionally, when the 10 lipids with the smallest P values in the training group were used directly as predictive features, the AUCs were 0.815 (95% CI: 0.724–0.906) in the training group and 0.749 (95% CI: 0.627–0.870) in the validation group, again showing inferior performance compared to the LASSO-selected panel. The ROC curves for all methods in both cohorts are presented in Fig. 4C and D.

Based on comparative performance, the six lipids selected through LASSO regression were ultimately adopted as the predictive panel. Among them, only PS (18:0_24:0) was upregulated in the Immu + Che-NR group, which exhibited poor response to combination immunotherapy and chemotherapy, while Cer (d18:1/26:0), Cer (d18:2/25:0), SM (d18:1/20:1), Cer (d17:1/25:1), and DG (O-18:1_16:0) were downregulated (Fig. 4E). A heatmap of these 6 lipids also demonstrated this expression pattern, with consistent downregulation of 5 lipids and upregulation of PS (18:0_24:0) in the Immu + Che-NR group (Fig. 4F).

Construction of a prognostic model

A prognostic model was then developed based on five differential metabolites. Using this model, PFS was evaluated by stratifying ES-SCLC patients who received first-line immunotherapy combined with chemotherapy into high-risk and low-risk groups according to their risk scores (Fig. 5A). Similarly, a prognostic model incorporating the 6 LASSO-selected lipids was used to categorize patients into risk groups and assess the relationship between lipidomic profiles and survival outcomes (Fig. 5B). Kaplan-Meier analysis showed that the median PFS for the low-risk and high-risk groups, based on metabolite profiles, was 11.0 months (95% CI: 9.5–12.5) and 7.0 months (95% CI: 6.2–7.8), respectively (P < 0.001), with significantly longer PFS observed in the low-risk group (HR: 0.548; 95% CI: 0.382–0.788; P = 0.001; Fig. 5C). Similarly, based on lipidomic profiles, the median PFS was 10.0 months (95% CI: 8.1–11.9) in the low-risk group and 6.0 months (95% CI: 5.1–6.9) in the high-risk group (P < 0.001), with significantly improved outcomes in the low-risk group (HR: 0.443; 95% CI: 0.304–0.644; P < 0.001; Fig. 5D).

Fig. 5 — Predictive effect of the constructed metabolite and lipid models on patient survival outcomes. (A) Triple plot of risk scores based on metabolite prediction model. Patients were divided into low risk and high risk groups using the median of the risk score as the cutoff value. (B) Triple plot of risk scores based on lipid prediction model. (C) Kaplan-Meier (KM) survival curves based on metabolite prediction models. (D) KM survival curves based on lipid prediction models. (E) Forest plot for multivariate Cox analysis. (F) KM survival curves for metabolite prediction models predicting prognosis in second-line therapy. (G) KM survival curves for lipid prediction models predicting prognosis in second-line therapy. TPS, tumor proportion score; BMI, body mass index; AEs, adverse events; Cer, ceramide; SM, sphingomyelin; DG, diacylglycerol; PS, Phosphatidylserine

A multivariate Cox regression model was constructed including clinical variables such as age, sex, smoking history, ECOG performance status, T stage, N stage, M stage, radiotherapy, adverse events, PD-L1 expression, body mass index, coronary heart disease, diabetes, hypertension, and the metabolite and lipid-based risk scores. Only the risk scores derived from metabolites and lipids were independently associated with PFS (Fig. 5E). Specifically, the metabolite low-risk group had significantly longer PFS compared to the high-risk group (adjusted HR: 0.596; 95% CI: 0.387–0.919; P = 0.019), and the lipid low-risk group also showed superior outcomes (adjusted HR: 0.489; 95% CI: 0.323–0.739; P = 0.001).

Additionally, among the 144 patients included in this analysis, 60 patients who experienced progression following first-line immunotherapy plus chemotherapy continued with second-line immunotherapy combined with chemotherapy and anlotinib. Notably, the prognostic models developed based on first-line treatment data also successfully predicted outcomes for these patients. Based on the metabolite-derived risk score, the median PFS for the low-risk and high-risk groups was 5.0 months (95% CI: 4.4–5.6) and 3.0 months (95% CI: 2.6–3.4), respectively (P = 0.009) (Fig. 5F). For the lipid-derived risk score, the median PFS was 5.0 months (95% CI: 4.3–5.7) in the low-risk group and 3.0 months (95% CI: 2.4–3.6) in the high-risk group (P = 0.008) (Fig. 5G). In both models, PFS was significantly prolonged in the low-risk group (HR: 0.565; 95% CI: 0.334–0.957; P = 0.034 for metabolites; HR: 0.558; 95% CI: 0.329–0.947; P = 0.031 for lipids).

Proteomics analysis and integrated multi-omics analysis

To explore the potential mechanisms by which specific metabolite and lipid combinations predict the prognosis of combined first- and second-line immunotherapy, we performed ultrafast blood proteomic analysis in patients receiving combined immunotherapy. Metabolomic data had previously revealed alterations in small-molecule metabolites, which we aimed to validate and further elucidate through proteomic profiling.

These patients were stratified into the Immu + che-R and Immu + che-NR groups based on PFS. As shown in Fig. 6A, the PCA plot demonstrated clear separation between the two groups along principal component 1 (PC1), which accounted for 15.41% of the total variance, despite some overlap between individual samples. Using the criteria of FC ≥ 1.5 or ≤ 0.6667 and P-value < 0.05, we initially identified 165 differentially expressed proteins. Among these, 29 proteins were upregulated and 136 were downregulated in the Immu + che-R group (Fig. 6B). Further screening of differential proteins was performed using Lasso regression, random forest algorithms, and ranking based on the ten most statistically significant P-values. The Lasso regression method identified six proteins—ENTPD5, LDHA, SOD3, SNRPF, SHH, and ADAMTS8—as a prognostic combination (Fig. 6C). The random forest method yielded three key proteins—ACIN1, ACSL4, and PHGDH—as another prognostic signature.

As shown in Fig. 6D, the ROC curve analysis revealed that the AUC for the combinations derived from Lasso regression, random forest, and the top 10 differential proteins were 0.912 (95% CI: 0.842–0.982), 0.935 (95% CI: 0.877–0.993), and 0.987 (95% CI: 0.965–1.000), respectively. All three combinations exhibited strong predictive performance. Notably, the random forest-derived combination demonstrated the highest clinical applicability. Therefore, ACIN1, ACSL4, and PHGDH were selected as the protein-based model to predict immunotherapy prognosis. Violin plots (Fig. 6E) indicated that the expression levels of ACIN1, ACSL4, and PHGDH were significantly lower in patients with poor prognosis, a trend further corroborated by the heatmap (Fig. 6F). KEGG pathway enrichment analysis revealed that these differential proteins were predominantly involved in cancer-related metabolic pathways, including the IL-17 signaling pathway, glycine, serine and threonine metabolism, and Th17 cell differentiation (Fig. 6G).

Patients were further classified into high-risk and low-risk groups based on the risk score calculated from the differential protein panel (Fig. 6H). During the first-line immunotherapy combined with chemotherapy stage, the median PFS was 10.0 months (95% CI: 8.7–11.3) for the low-risk group, compared with 6.0 months (95% CI: 4.5–7.5) for the high-risk group (P < 0.001). During the second-line immunotherapy plus chemotherapy and anlotinib stage, the median PFS was 5.0 months (95% CI: 4.6–5.4) in the low-risk group and 3.0 months (95% CI: 2.5–3.5) in the high-risk group (P = 0.002) (Fig. 6I). PFS was significantly longer in the low-risk group than in the high-risk group in both treatment stages (first-line: HR = 0.445, 95% CI: 0.261–0.758, P = 0.003; second-line: HR = 0.513, 95% CI: 0.303–0.869, P = 0.013).

Subsequently, we conducted an integrated multi-omics analysis to identify KEGG pathways jointly enriched in both metabolomics and proteomics, as well as in lipidomics and proteomics. We observed that all three omics layers were consistently enriched in the glycine, serine, and threonine metabolism pathway (Fig. 6J). PHGDH was identified as the differential protein involved in this pathway, while L-aspartate-semialdehyde and L-cysteine were the associated metabolites, and PS(18:0_24:0) was the corresponding lipid. All of these biomarkers had been included in our previously established prognostic models for immunotherapy combined with chemotherapy.

As shown in Fig. 6K, correlation analyses revealed that PHGDH exhibited positive associations with L-aspartate-semialdehyde, L-cysteine, and PS(18:0_24:0). These findings suggest that glycine, serine, and threonine metabolism undergo coordinated alterations at the metabolite, lipid, and protein levels. We speculate that the upregulation of PHGDH enhances the production of key metabolites (L-aspartate-semialdehyde and L-cysteine) and lipids (PS(18:0_24:0)) in this metabolic pathway, thereby diminishing the efficacy of immunotherapy combined with chemotherapy in patients with ES-SCLC and potentially impacting the effectiveness of anlotinib as well.

Serum multi-omics profile after one cycle of treatment

To dynamically monitor the effect of changes in metabolites, lipids, and proteins during treatment cycles on predicting the prognosis of immunotherapy, we also compared the differences between NR and R groups after receiving immunotherapy combined with chemotherapy for one cycle. As shown in Fig. 7A, the 5 metabolites in the previous metabolite prediction model were still significantly different: Uric acid and Dimethisterone were significantly downregulated, while L-Aspartate-semialdehyde, Xanthine, and L-Cysteine were upregulated in patients who showed poor response to combined immunotherapy. Similarly, the 6 lipids and 3 proteins previously shown to be significantly different continue to show the same trend in Fig. 7B and C.

Fig. 7 — Metabolomics, lipidomics and proteomics analysis of patients after one cycle of treatment. (A) Volcano plot of differential metabolites for Immu + che-R and Immu + che NR groups after one cycle of treatment. (B) Volcano plot of differential lipids for Immu + che-R and Immu + che NR groups after one cycle of treatment. (C) Volcano plot of differential proteins for Immu + che-R and Immu + che NR groups after one cycle of treatment. (D) Relative amounts of metabolites in the metabolite prognostic model in the Immu + che-R and Immu + che NR groups at baseline and after one cycle of treatment, respectively. (E) Relative amounts of proteins in the protein prognostic model in the Immu + che-R and Immu + che NR groups at baseline and after one cycle of treatment, respectively. (F) Relative amounts of lipids in the lipid prognostic model in the Immu + che-R and Immu + che NR groups at baseline and after one cycle of treatment, respectively. (G) Kaplan-Meier (KM) survival curves for lipid models after one cycle of therapy predicting prognosis. (H) Kaplan-Meier (KM) survival curves for protein models after one cycle of therapy predicting prognosis. R, response; NR, non-response; C1-R, response after one cycle of treatment; C1-NR, non-response after one cycle of treatment; ACIN1, apoptotic chromatin condensation inducer 1; ACSL4, long-chain acyl-CoA synthetase 4; PHGDH, phosphoglycerate dehydrogenase; Cer, ceramide; SM, sphingomyelin; DG, diacylglycerol; PS, Phosphatidylserine

In addition, we compared the relative content changes of these metabolites, lipids, and proteins between the R and NR groups at baseline and after one cycle of treatment. Figure 7D showed the changes in 5 metabolites, Uric acid and Dimethisterone were significantly upregulated in R group at both baseline and after one cycle of treatment compared to NR group, while the relative content of Dimethisterone was significantly upregulated from baseline after one cycle of treatment. Compared with NR group, L-Aspartate-semialdehyde, Xanthine, and L-Cysteine were significantly down-regulated in R group at baseline and after one cycle of treatment, and the contents of these metabolites were significantly down-regulated after one cycle of treatment. Figure 7E showed the changes in three proteins, ACIN1, ACSL4, and PHGDH were significantly down-regulated in R group at both baseline and after one cycle of treatment compared with NR group, while the relative content of PHGDH showed a trend of continued down-regulation from baseline after one cycle of treatment. Figure 7F showed the changes in 6 lipids, Cer (d18:1/26:0), Cer (d18:2/25:0), SM (d18:1/20:1), Cer (d17:1/25:1), and DG (O-18:1_16:0) were significantly upregulated in R group at baseline and after one cycle of treatment compared with NR group, and the relative contents of SM (d18:1/20:1) and DG (O-18:1_16:0) continued to rise from baseline after one cycle of treatment. Compared with NR group, PS (18:0_24:0) was significantly down-regulated in R group at baseline and after one cycle of treatment, and the relative content continued to decrease from baseline after one cycle of treatment.

We were also curious whether the relative content of these metabolites, lipids, and protein combinations after one cycle of treatment could predict the prognosis of patients, so we divided patients into high risk and low risk groups according to the median relative content of the combinations. We found that patients in the high risk group had significantly shorter PFS than those in the low risk group in the metabolite-related survival curves shown in Fig. 7G (P = 0.027). Additionally, in the lipid-associated survival curve shown in Fig. 7H, PFS was significantly shorter in the high risk group than in the low risk group (P = 0.008). Similarly, in the protein-related survival curve shown in Fig. 7I, PFS was significantly shorter in patients in the high risk group than in those in the low risk group (P = 0.029). This suggested that detecting the content of these metabolites, lipids, and proteins in patients after one cycle of immunotherapy combined with chemotherapy could still predict the prognosis of these patients.

Discussion

In this study, we developed 5 metabolite combinations, 6 lipid combinations, and 3 protein combinations as potential prognostic biomarkers for first-line immunotherapy combined with chemotherapy in ES-SCLC. These combinations were identified through metabolomic, lipidomic, and proteomic analyses of serum samples from 816 samples, in conjunction with machine learning methods. Notably, these biomarker panels also demonstrated predictive value for prognosis in the context of second-line immunotherapy combined with chemotherapy and anlotinib. Furthermore, integrated analysis of metabolomics, lipidomics, and proteomics revealed that the glycine, serine, and threonine metabolic pathway may play a pivotal role in influencing treatment outcomes in ES-SCLC patients receiving first-line immunotherapy combined with chemotherapy. Our findings suggested that PHGDH was upregulated in patients with poor prognosis, leading to increased activity in this metabolic pathway and elevated levels of its downstream products—L-cysteine and PS(18:0_24:0). Detecting metabolite, lipid and protein levels after 1 cycle of treatment could still accurately predict prognosis, and the expression trends of these markers remained consistent before and after treatment, with real-time dynamic monitoring potential.

It is well established that the prognosis of patients with SCLC is generally poor due to the aggressive nature of the disease, rapid progression, and limited treatment options. Furthermore, the incidence of SCLC has been rising in recent years, with most cases being diagnosed at an advanced stage [27, 28]. Although the introduction of first-line immunotherapy combined with chemotherapy has modestly improved outcomes in patients with ES-SCLC, biomarkers predictive of immunotherapy response remain under investigation [29]. In the present study, we employed a multi-omics approach, simultaneously analyzing the proteome, metabolome, and lipidome, to identify potential biomarkers associated with the efficacy of combined immunotherapy and chemotherapy in ES-SCLC. Our findings, based on a large cohort of ES-SCLC patients, demonstrated that first-line immunotherapy combined with chemotherapy resulted in improved prognosis compared with chemotherapy alone. Unlike tumor tissues, serum samples are non-invasive and readily obtainable, allowing for dynamic monitoring during treatment. Compared with single-omics strategies, multi-omics approaches offer a more comprehensive perspective and enable deeper insight into the mechanisms underlying drug resistance and therapeutic benefit.

Metabolomics is particularly effective in detecting polar metabolites such as amino acids, organic acids, and sugars [30], whereas lipidomics specializes in the profiling of non-polar lipids, including phospholipids, sphingolipids, and glycerides [31]. The combined application of metabolomics and lipidomics provides a more holistic overview of blood metabolic alterations. Although limited studies have applied lipidomics in lung cancer research, lipids are not only essential components of cell membranes but also serve as key mediators of signal transduction and immune regulation. Lipidomics facilitates the characterization of lipid exchange between immune cells (e.g., CD8⁺ T cells, tumor-associated macrophages) and tumor cells, thereby revealing dynamic changes in the tumor microenvironment associated with immunosuppression or activation [32]. We therefore propose that lipidomic profiling plays a critical role in identifying prognostic biomarkers for immunotherapy in ES-SCLC. To date, lipidomics has predominantly been applied in NSCLC, where lipid markers such as PC (16:0/18:2) and CE (20:1) have been identified to predict the efficacy of combined immunotherapy and chemotherapy, with the predictive model achieving an AUC of 0.87 [33]. However, applications of lipidomics in SCLC remain scarce. Our study aims to fill this gap.

By systematically integrating serum-based metabolomic, lipidomic, and proteomic data, we constructed biomarker models to predict the efficacy of first-line immunotherapy combined with chemotherapy in ES-SCLC, utilizing machine learning algorithms. Initially, differential lipids were screened in the overall cohort. Subsequently, robust metabolite and lipid biomarkers were identified through training and validation cohorts, demonstrating excellent predictive performance in both sets. Further risk score modeling and multivariate Cox regression analysis confirmed that these biomarker combinations were significantly associated with the prognosis of patients receiving first-line immunotherapy combined with chemotherapy. Notably, these biomarker combinations also showed potential in predicting the efficacy of second-line treatment regimens involving anlotinib. This finding provides fresh evidence for exploring how anti-angiogenic and immune therapies work together.

A small retrospective study reported that a second-line regimen combining anlotinib with PD-1/PD-L1 inhibitors achieved a median PFS of 8.2 months and OS of 20.1 months in 26 ES-SCLC patients, outperforming outcomes from single-center studies of anlotinib combined with chemotherapy alone (PFS of 5.6 months and OS of 15.1 months) [34]. However, due to its non-randomized, single-arm design, the study does not provide conclusive evidence for a true synergistic effect between anlotinib and immune checkpoint inhibitors. Experimental studies have shown that anlotinib, by targeting VEGFR, FGFR, and c-Kit, can promote vascular normalization in tumors, enhance CD8⁺ T cell infiltration, and reduce the numbers of myeloid-derived suppressor cells (MDSCs) and regulatory T cells (Tregs), thereby establishing a theoretical basis for “anti-angiogenic plus immunotherapy” combination strategies. Our study may contribute to the development of non-invasive, dynamic tools for predicting treatment efficacy, facilitating the identification of patients most likely to benefit from such combination therapies, and advancing understanding of their underlying synergistic mechanisms [25, 35].

To further elucidate the mechanisms responsible for differences in therapeutic efficacy, we conducted a proteomic analysis of serum samples from this ES-SCLC patient cohort. The protein-based prediction model, developed using machine learning algorithms, successfully distinguished responders from non-responders and demonstrated high prognostic accuracy in both training and validation cohorts (AUC > 0.85). Integrated KEGG pathway analysis across multi-omics datasets and Spearman correlation analysis consistently highlighted the glycine–serine–threonine metabolic axis. Key proteins (PHGDH), core metabolites (L-aspartate-semialdehyde, L-cysteine), and characteristic lipids [PS (18:0_24:0)] involved in this pathway were all identified within our previously established prognostic model and exhibited significant positive correlations with one another, further supporting their prognostic relevance in the context of immunotherapy.

The glycine–serine–threonine metabolic axis serves as a central hub integrating cellular one-carbon metabolism with nucleotide, glutathione (GSH), and phospholipid biosynthesis [36]. PHGDH (3-phosphoglycerate dehydrogenase), the rate-limiting enzyme in this pathway, catalyzes the conversion of the glycolytic intermediate 3-phosphoglycerate (3-PG) to 3-phosphohydroxypyruvate (3-PHP), which subsequently contributes to serine biosynthesis [37, 38]. Cells with elevated PHGDH expression divert substantial carbon flux from glycolysis and the tricarboxylic acid (TCA) cycle into this pathway, thereby supporting sustained glycolytic flux and ATP production. Simultaneously, this shunt generates large quantities of one-carbon donors, such as 5,10-methylenetetrahydrofolate, which accelerate purine and pyrimidine synthesis, thus promoting rapid tumor proliferation [39]. L-aspartate-semialdehyde, located at the intersection of this axis with lysine and threonine biosynthesis, may be reduced to homoserine or further oxidized to oxaloacetate, replenishing the TCA cycle and contributing to metabolic redundancy, which maintains energy homeostasis under immune pressure [40]. L-cysteine, through the transsulfuration pathway, contributes to the biosynthesis of GSH. Elevated GSH levels scavenge reactive oxygen species (ROS) induced by immuno-chemotherapy, protect tumor cells from apoptosis, and inhibit dendritic cell maturation. Therefore, the concurrent elevation of L-aspartate-semialdehyde and L-cysteine suggests a dual advantage of enhanced proliferation and antioxidative capacity [41, 42].

The glycine–serine–threonine metabolic axis drives the production and function of phosphatidylserine PS (18:0_24:0) through two complementary circuits, which, in turn, influence the response of ES-SCLC to immune combination therapy [43]. The first circuit is the “substrate supply pathway”: under hypoxic conditions, tumor cells initiate the Warburg effect, wherein the glycolytic intermediate 3-PG is continuously catalyzed by the highly expressed PHGDH to generate serine [44]. This serine serves dual roles—directly acting as a substrate for PTDSS1/2 to convert PC or PE into PS (18:0_24:0) at the membrane, and reversibly interconverting with glycine via SHMT to form the “serine–glycine–one-carbon unit” cycle, which provides continuous methyl donors (SAM) for nucleotide synthesis and DNA repair [45, 46]. Threonine further supports this cycle by being cleaved via threonine aldolase into glycine and acetyl-CoA [47]. Thus, PHGDH upregulation not only expands the intracellular serine pool but also indirectly facilitates PS (18:0_24:0) synthesis, offering proliferative and membrane signaling advantages to tumor cells.

Excess PS (18:0_24:0), evaginated to the outer leaflet of the plasma membrane, acts as a “don’t-eat-me” signal and synergizes with immune checkpoints to inhibit dendritic cell maturation and CD8⁺ T cell infiltration [48]. Concurrently, serine/glycine metabolic reprogramming reduces intracellular ROS levels and suppresses immunogenic cell death [49]. We observed that PHGDH, PS (18:0_24:0), and its downstream metabolite L-cysteine were significantly elevated in the serum of non-responders to immune combination therapy, whereas these markers were downregulated in responders. These findings suggest that inhibiting PHGDH or blocking PS eversion could reverse the immunosuppressive tumor microenvironment and enhance immunotherapeutic efficacy. Therefore, combining agents that target this metabolic axis (e.g., PHGDH inhibitors, PS eversion inhibitors) with PD-1/PD-L1 inhibitors may represent a novel therapeutic strategy to overcome primary resistance in ES-SCLC.

Previous studies have demonstrated that glycine, serine, and threonine metabolism play critical roles in the initiation, progression, therapeutic response, and immune regulation of NSCLC. For instance, it has been reported that lung cancer cells significantly upregulate serine–glycine metabolism following radiotherapy to facilitate DNA repair and promote survival, a capacity not observed in normal lung tissue. Inhibiting this metabolic pathway—such as through the administration of sertraline—has been shown to markedly enhance radiotherapeutic efficacy in NSCLC, suppress tumor growth, and improve the immune microenvironment [50]. Similarly, genetic inhibition of IDH1 or pharmacological elimination of agents that support serine metabolism have been shown to impede NSCLC progression and enhance the antitumor effects of gemcitabine and serine–glycine starvation therapies [51]. However, no prior studies have established a direct functional or predictive link between glycine, serine, and threonine metabolism and the therapeutic efficacy (e.g., PFS, ORR) of immunotherapy or combination immunotherapy in patients with ES-SCLC.

In our study, integrated multi-omics analyses consistently identified alterations in the glycine–serine–threonine metabolic pathway at the protein, metabolite, and lipid levels, strongly indicating that this axis constitutes a core mechanism underlying prognostic divergence in response to combination immunochemotherapy. Notably, key components of this pathway—namely the enzyme PHGDH, the metabolites L-aspartate-semialdehyde and L-cysteine, and the lipid PS (18:0_24:0)—were all included in our previously constructed prognostic model, and demonstrated strong positive correlations with one another. Moreover, we detected metabolites, lipids, and proteins in serum samples from patients treated for one cycle and found that our established prediction model could still accurately predict the prognosis of patients. Among them, the expression trend of key markers remains consistent before and after treatment and has the potential for real-time dynamic monitoring. Compared with baseline, L-aspartate-semialdehyde and L-cysteine, PS (18:0_24:0) and PHGDH decreased further after treatment, suggesting that perhaps the glycine, serine, threonine metabolic pathway also continued to function during treatment. These cross-omics concordance findings not only underscore the central role of this metabolic pathway in modulating treatment outcomes but also provide verifiable molecular targets for subsequent mechanistic studies. ALL in all, our study targeted first-line immunotherapy plus chemotherapy SCLC patients and provided a combined marker panel consisting of 5 metabolites, 6 lipids, and 3 proteins that could stratify high-risk patients by routine serum testing. This model is equally effective in second-line anlotinib combined with immunochemotherapy, providing a rapid prognostic assessment tool for patients who have progressed on first-line therapy and helping clinicians to individualize subsequent treatment regimens. Biomarkers detection requires only peripheral blood samples, is non-invasive and easy to obtain, is suitable for widespread promotion, and solves the clinical problem of ES-SCLC lacking effective prognostic markers. Dynamic monitoring also facilitates timely adjustment of treatment regimens, and prognosis can be predicted by detecting core marker levels after 1 cycle of treatment, breaking through the lag limitation of traditional reliance on imaging assessment and achieving early prediction of treatment outcomes. Core markers (such as PHGDH, L-aspartate semialdehyde, PS (18:0_24:0), etc.) have consistent expression trends before and after treatment and can be used as dynamic monitoring indicators to reflect the changes in the response of patients to treatment in real time and guide the timely adjustment of clinical treatment strategies.

PD-L1 is already a well-established biomarker in NSCLC immunotherapy, but its predictive value in ES-SCLC is significantly limited. PD-L1 expression rates in tumor tissue are generally not high in ES-SCLC, and neither CASPIAN nor IMpower133 ^{14, 15} have found an association between PD-L1 levels and efficacy. Plasma soluble PD-L1 is less specific and increases tend to reflect systemic inflammation rather than tumor-specific immunosuppression. In contrast, our multi-omics panel (5 metabolites, 6 lipids, 3 proteins) maintained high predictive accuracy in the entire PD-L1 subgroup, clearly distinguishing high and low risk even in patients with TPS < 1%, filling the marker-free gap in populations with low PD-L1 expression. Unlike PD-L1, which only assesses tumor cell autoantigen presentation, our integrated model simultaneously quantifies metabolic reprogramming and immune microenvironment regulation, is more convenient and dynamic monitoring than tissue PD-L1, more specific than peripheral blood PD-L1, and provides a more comprehensive systematic perspective of treatment response. TMB has been shown to predict immune checkpoint inhibitor efficacy in a variety of solid tumors, but both phase III studies of CASPIAN and IMpower133 ^{14, 15} showed no significant association between tissue TMB and PFS/OS, possibly due to the low SCLC mutation load, bTMB is even less relevant literature to support its predictive role in SCLC, and there is no uniform threshold for ES-SCLC. While our serum multi-omics panel maintained high predictive performance in any M stage, smoking status and other subgroups, the experimental process only required conventional LC-MS/MS and machine learning modeling, with controllable costs and short cycles. Mechanistically, the Panel focuses on the glycine-serine-threonine metabolic axis and directly captures the core metabolic links of tumor immune interactions; while TMB enumerates only genetic mutations, which are limited in information content in low-mutant SCLC and difficult to provide equivalent functional insight. CtDNA is a promising liquid biopsy tool for monitoring immune response, but it is still insufficient in predicting immune efficacy in ES-SCLC. The baseline detection rate was only 60–70% and was more difficult to detect in patients with low tumor burden. It relies on tumor-specific mutations, while the SCLC mutation spectrum is highly dispersed and it is difficult to simultaneously capture the metabolic and immune dynamics required for the immune response. While our serum markers detected target metabolites, lipids, and proteins in all 144 baseline samples from ES-SCLC patients who received immunotherapy plus chemotherapy; samples were retaken after completion of 1 cycle of treatment, the Panel maintained a high AUC and the expression trend was consistent before and after, without waiting for slow changes in the frequency of multiple cycles of mutant alleles as ctDNA. More importantly, Panel reveals the glycine-serine-threonine metabolic reprogramming mechanism through core molecules such as PHGDH, while ctDNA only indirectly reflects tumor burden and cannot provide functional information at the same depth.

Despite these novel findings, our study has several limitations. First, it was a retrospective single-center study, which may introduce unavoidable biases. Nevertheless, we included as many patients as possible, ensured baseline comparability, and attempted to minimize bias. We anticipate that future prospective, multicenter studies will be conducted to validate our hypothesis. Second, the study lacked external validation. Although we established discovery, training, and validation cohorts to improve the reliability of our findings, validation using an external cohort or experimental models is still warranted. Lastly, the lack of OS analysis is a limitation due to immature OS data and loss to follow-up in some cases; however, we expect to obtain more robust OS data for future statistical analyses. We will try our best to perfect our study from three aspects, firstly, combine multiple tumor research centers to further expand the sample size of SCLC. Secondly, increase basic experimental research to explore the mechanism of the occurrence and development of the results of this study. Thirdly, the addition of metabolic and immune-related studies on other lung cancer types, such as NSCLC.

In conclusion, we systematically identify prognostic biomarkers from protein, metabolite, and lipid levels that are significantly associated with the efficacy and prognosis of ES-SCLC immunochemotherapy, providing a comprehensive characterization of protein-metabolite signatures in peripheral blood of patients with different prognoses. Based on these findings, we developed a multi-omics biomarker panel capable of accurately predicting the efficacy of both first-line immunochemotherapy and second-line combination therapy with anlotinib in ES-SCLC. Crucially, integrated metabolomic, lipidomic, and proteomic data revealed that metabolic reprogramming of glycine, serine, and threonine constitutes the core mechanism driving therapeutic differences. PHGDH, the rate-limiting enzyme of this pathway, is markedly upregulated. This causes excess accumulation of L-aspartate-semialdehyde, L-cysteine and PS (18:0_24:0), metabolites linked to poor prognosis, and suggests that targeting the axis could overcome immunotherapy resistance. Moreover, dynamic monitoring of peripheral blood PHGDH, as well as its related metabolites and lipids, including L-aspartate-semialdehyde, L-cysteine, and PS (18:0_24:0), may enable real-time identification of patients likely to benefit from immunotherapy. This study offers translational biomarkers and therapeutic targets for precise stratification and for the development of metabolism–immunity combinatorial strategies aimed at improving the prognosis of ES-SCLC patients and overcoming current limitations in immunotherapy efficacy.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Material 1^{(1.3MB, docx)}

Acknowledgements

We thank Biostatisticians at the Statistical Center of Shanghai Chest Hospital for their help with our statistical analysis data.

Author contributions

Liang Zheng and Haoming Xu: Writing Original Draft; Shuyuan Wang and Meili Ma: Conceptualization; Fang Hu and Lei Cheng: Data curation; Jun Lu and Feng Pan: Figures; Bo Zhang and Jianlin Xu: Literature Search; Ying Li and Yinchen Shen: Resources; Wei Zhang, Runbo Zhong, and Tianqing Chu: Methodology; Baohui Han and Xiaoxuan Zheng: Project administration; Hua Zhong, Wei Nie and Xueyan Zhang: Writing review and editing. All authors contributed to the article and approved the submitted version.

Funding

This study was supported by Shanghai Innovative Medical Device Application Demonstration Project 2023 (NO. 23SHS02600); National Natural Science Foundation of China (No. 82373425 and No. 82400126); the Medical Innovation Research Special Project of the Science and Technology Commission of Shanghai Municipality (No. 23Y11904200); Noncommunicable Chronic Diseases-National Science and Technology Major Project (2024ZD0529300); Scientific Research Program by Shanghai Municipal Education Commission (2024AIYBO11); Clinical Research Special Fund Project of Shanghai Chest Hospital (2024IIT-M001); Technology Plan Projects in Zhejiang Province (NO.2025KY034); The NSFC Cultivation Project of Zhejiang Cancer Hospital (NO.PY2023021). The funder didn’t influence the results/outcomes of the study despite author affiliations with the funder.

Data availability

Data can obtain with the permission of the corresponding author.

Declarations

Ethical approval

The study complied with all relevant ethical regulations in accordance with the Declaration of Helsinki and was approved by the Ethics Committee and Institutional Review Board of Shanghai Chest Hospital (Reference number: LS1808). Written informed consent was obtained from all patients.

Consent for publication

Not applicable.

Generative AI and AI-assisted technologies

Generative AI and AI-assisted technologies were NOT used in the preparation of this work.

Competing interests

The authors state that they have no conflict of interests.

Footnotes

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Liang Zheng and Haoming Xu contributed equally to this work.

Contributor Information

Baohui Han, Email: 18930858216@163.com.

Xiaoxuan Zheng, Email: milozheng59@163.com.

Hua Zhong, Email: eddiedong8@hotmail.com.

Wei Nie, Email: niewei-1001@163.com.

Xueyan Zhang, Email: zxychest0109@163.com.

References

1.Yin X, Li Y, Wang H, et al. Small cell lung cancer transformation: from pathogenesis to treatment. Semin Cancer Biol Nov. 2022;86(Pt 2):595–606. 10.1016/j.semcancer.2022.03.006. [DOI] [PubMed] [Google Scholar]
2.Kim SY, Park HS, Chiang AC. Small cell lung cancer: A review. Jama Jun. 2025;3(21):1906–17. 10.1001/jama.2025.0560. [DOI] [PubMed] [Google Scholar]
3.Herzog BH, Devarakonda S, Govindan R. Overcoming chemotherapy resistance in SCLC. J Thorac Oncol Dec. 2021;16(12):2002–15. 10.1016/j.jtho.2021.07.018. [DOI] [PubMed] [Google Scholar]
4.Fischer B, Marinov M, Arcaro A. Targeting receptor tyrosine kinase signalling in small cell lung cancer (SCLC): what have we learned so far? Cancer Treat Rev Jun. 2007;33(4):391–406. 10.1016/j.ctrv.2007.01.006. [DOI] [PubMed] [Google Scholar]
5.Gomez-Randulfe I, Leporati R, Gupta B, Liu S, Califano R. Recent advances and future strategies in first-line treatment of ES-SCLC. Eur J Cancer Mar. 2024;200:113581. 10.1016/j.ejca.2024.113581. [DOI] [PubMed] [Google Scholar]
6.Cheng Y, Fan Y, Zhao Y, et al. Tislelizumab plus platinum and Etoposide versus placebo plus platinum and Etoposide as First-Line treatment for Extensive-Stage SCLC (RATIONALE-312): A Multicenter, Double-Blind, placebo-Controlled, Randomized, phase 3 clinical trial. J Thorac Oncol Jul. 2024;19(7):1073–85. 10.1016/j.jtho.2024.03.008. [DOI] [PubMed] [Google Scholar]
7.Chen D, Zou B, Li B, et al. Adebrelimab plus chemotherapy and sequential thoracic radiotherapy as first-line therapy for extensive-stage small-cell lung cancer (ES-SCLC): a phase II trial. EClinicalMedicine Sep. 2024;75:102795. 10.1016/j.eclinm.2024.102795. [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Xie M, Vuko M, Rodriguez-Canales J, et al. Molecular classification and biomarkers of outcome with immunotherapy in extensive-stage small-cell lung cancer: analyses of the CASPIAN phase 3 study. Mol Cancer May. 2024;30(1):115. 10.1186/s12943-024-02014-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Yi M, Niu M, Xu L, Luo S, Wu K. Regulation of PD-L1 expression in the tumor microenvironment. J Hematol Oncol Jan. 2021;7(1):10. 10.1186/s13045-020-01027-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Nabet BY, Hamidi H, Lee MC, et al. Immune heterogeneity in small-cell lung cancer and vulnerability to immune checkpoint Blockade. Cancer Cell Mar. 2024;11(3):429–e4434. 10.1016/j.ccell.2024.01.010. [DOI] [PubMed] [Google Scholar]
11.Paz-Ares L, Garassino MC, Chen Y, et al. Durvalumab ± Tremelimumab + Platinum-Etoposide in Extensive-Stage small cell lung cancer (CASPIAN): outcomes by PD-L1 expression and tissue tumor mutational burden. Clin Cancer Res Feb. 2024;16(4):824–35. 10.1158/1078-0432.Ccr-23-1689. [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Xue Y, Gao S, Gou J, et al. Platinum-based chemotherapy in combination with PD-1/PD-L1 inhibitors: preclinical and clinical studies and mechanism of action. Expert Opin Drug Deliv Feb. 2021;18(2):187–203. 10.1080/17425247.2021.1825376. [DOI] [PubMed] [Google Scholar]
13.Chen T, Wang M, Chen Y, Cao Y, Liu Y. Advances in predictive biomarkers associated with immunotherapy in extensive-stage small cell lung cancer. Cell Biosci Sep. 2024;12(1):117. 10.1186/s13578-024-01283-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Paz-Ares L, Dvorkin M, Chen Y, et al. Durvalumab plus platinum-etoposide versus platinum-etoposide in first-line treatment of extensive-stage small-cell lung cancer (CASPIAN): a randomised, controlled, open-label, phase 3 trial. Lancet Nov. 2019;23(10212):1929–39. 10.1016/s0140-6736(19)32222-6. [DOI] [PubMed] [Google Scholar]
15.Liu SV, Reck M, Mansfield AS, et al. Updated overall survival and PD-L1 subgroup analysis of patients with Extensive-Stage Small-Cell lung cancer treated with Atezolizumab, Carboplatin, and Etoposide (IMpower133). J Clin Oncol Feb. 2021;20(6):619–30. 10.1200/jco.20.01055. [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Killock D. Prospective insights on the use of bTMB. Nat Rev Clin Oncol Jun. 2022;19(6):360. 10.1038/s41571-022-00640-2. [DOI] [PubMed] [Google Scholar]
17.Xu Z, Zeng S, Gong Z, Yan Y. Exosome-based immunotherapy: a promising approach for cancer treatment. Mol Cancer Nov. 2020;12(1):160. 10.1186/s12943-020-01278-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Stadler JC, Belloum Y, Deitert B, et al. Current and future clinical applications of ctdna in Immuno-Oncology. Cancer Res Feb. 2022;1(3):349–58. 10.1158/0008-5472.Can-21-1718. [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Yi K, Wang X, Filippov SK, Zhang H. Emerging ctdna detection strategies in clinical cancer theranostics. Smart Med Nov. 2023;2(4):e20230031. 10.1002/smmd.20230031. [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Huo Z, Duan Y, Zhan D, et al. Proteomic stratification of prognosis and treatment options for small cell lung cancer. Genom Proteom Bioinf. Jul 2024;3(2). 10.1093/gpbjnl/qzae033. [DOI] [PMC free article] [PubMed]
21.Solta A, Ernhofer B, Boettiger K, et al. Small cells - big issues: biological implications and preclinical advancements in small cell lung cancer. Mol Cancer Feb. 2024;24(1):41. 10.1186/s12943-024-01953-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Chen H, Deng C, Gao J, et al. Integrative Spatial analysis reveals tumor heterogeneity and immune colony niche related to clinical outcomes in small cell lung cancer. Cancer Cell Mar. 2025;10(3):519–e5365. 10.1016/j.ccell.2025.01.012. [DOI] [PubMed] [Google Scholar]
23.Liu Q, Zhang J, Guo C, et al. Proteogenomic characterization of small cell lung cancer identifies biological insights and subtype-specific therapeutic strategies. Cell Jan. 2024;4(1):184–e20328. 10.1016/j.cell.2023.12.004. [DOI] [PubMed] [Google Scholar]
24.Cheng Y, Chen J, Zhang W, et al. Benmelstobart, anlotinib and chemotherapy in extensive-stage small-cell lung cancer: a randomized phase 3 trial. Nat Med Oct. 2024;30(10):2967–76. 10.1038/s41591-024-03132-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Shen G, Zheng F, Ren D, et al. Anlotinib: a novel multi-targeting tyrosine kinase inhibitor in clinical development. J Hematol Oncol Sep. 2018;19(1):120. 10.1186/s13045-018-0664-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Han X, Guo J, Li L, et al. Sintilimab combined with anlotinib and chemotherapy as second-line or later therapy in extensive-stage small cell lung cancer: a phase II clinical trial. Signal Transduct Target Ther Sep. 2024;16(1):241. 10.1038/s41392-024-01957-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Sen T, Takahashi N, Chakraborty S, et al. Emerging advances in defining the molecular and therapeutic landscape of small-cell lung cancer. Nat Rev Clin Oncol Aug. 2024;21(8):610–27. 10.1038/s41571-024-00914-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Thai AA, Solomon BJ, Sequist LV, Gainor JF, Heist RS. Lung cancer. Lancet. Aug 7. 2021;398(10299):535–554. 10.1016/s0140-6736(21)00312-3 [DOI] [PubMed]
29.Hendriks LE, Kerr KM, Menis J, et al. Oncogene-addicted metastatic non-small-cell lung cancer: ESMO clinical practice guideline for diagnosis, treatment and follow-up. Ann Oncol Apr. 2023;34(4):339–57. 10.1016/j.annonc.2022.12.009. [DOI] [PubMed] [Google Scholar]
30.Bauermeister A, Mannochio-Russo H, Costa-Lotufo LV, Jarmusch AK, Dorrestein PC. Mass spectrometry-based metabolomics in Microbiome investigations. Nat Rev Microbiol Mar. 2022;20(3):143–60. 10.1038/s41579-021-00621-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Wu Z, Bagarolo GI, Thoröe-Boveleth S, Jankowski J. Lipidomics: mass spectrometric and chemometric analyses of lipids. Adv Drug Deliv Rev. 2020;159:294–307. 10.1016/j.addr.2020.06.009. [DOI] [PubMed] [Google Scholar]
32.Wang R, Li B, Lam SM, Shui G. Integration of lipidomics and metabolomics for in-depth Understanding of cellular mechanism and disease progression. J Genet Genomics Feb. 2020;20(2):69–83. 10.1016/j.jgg.2019.11.009. [DOI] [PubMed] [Google Scholar]
33.Yu J, Xu H, Xiong F, et al. Lipidomics reveals biomarkers of the efficacy of first-line ICI therapy combined with chemotherapy in NSCLC. J Transl Med Jun. 2025;10(1):638. 10.1186/s12967-025-06542-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Xiong J, Xia L. Efficacy and safety of anlotinib as maintenance treatment in extensive-stage small cell lung cancer: a single-armed single center retrospective study. Front Oncol. 2024;14:1462581. 10.3389/fonc.2024.1462581. [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Han B, Li K, Wang Q, et al. Effect of anlotinib as a Third-Line or further treatment on overall survival of patients with advanced Non-Small cell lung cancer: the ALTER 0303 phase 3 randomized clinical trial. JAMA Oncol Nov. 2018;1(11):1569–75. 10.1001/jamaoncol.2018.3039. [DOI] [PMC free article] [PubMed] [Google Scholar]
36.Xue MY, Sun HZ, Wu XH, Liu JX, Guan LL. Multi-omics reveals that the rumen microbiome and its metabolome together with the host metabolome contribute to individualized dairy cow performance. Microbiome. May 12. 2020;8(1):64. 10.1186/s40168-020-00819-8 [DOI] [PMC free article] [PubMed]
37.Rossi M, Altea-Manzano P, Demicco M, et al. PHGDH heterogeneity potentiates cancer cell dissemination and metastasis. Nature May. 2022;605(7911):747–53. 10.1038/s41586-022-04758-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Zhang D, Li AM, Hu G, et al. PHGDH-mediated endothelial metabolism drives glioblastoma resistance to chimeric antigen receptor T cell immunotherapy. Cell Metab Mar. 2023;7(3):517–e5348. 10.1016/j.cmet.2023.01.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
39.Li SR, Bu LL, Cai L. Cuproptosis: lipoylated TCA cycle proteins-mediated novel cell death pathway. Signal Transduct Target Ther May. 2022;13(1):158. 10.1038/s41392-022-01014-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Zhan D, Wang D, Min W, Han W. Exploring the molecular basis for selective binding of Homoserine dehydrogenase from Mycobacterium leprae TN toward inhibitors: a virtual screening study. Int J Mol Sci Jan. 2014;24(2):1826–41. 10.3390/ijms15021826. [DOI] [PMC free article] [PubMed] [Google Scholar]
41.Mitchell SC, Waring RH. The deficiency of sulfoxidation of S-carboxymethyl-L-cysteine. Pharmacol Ther. 1989;43(2):237–49. 10.1016/0163-7258(89)90120-4. [DOI] [PubMed] [Google Scholar]
42.Pokorný V, Štejfa V, Havlín J, Fulem M, Růžička K. Heat capacities of L-Cysteine, L-Serine, L-Threonine, L-Lysine, and L-Methionine. Molecules Jan. 2023;3(1). 10.3390/molecules28010451. [DOI] [PMC free article] [PubMed]
43.Vanhove K, Graulus GJ, Mesotten L, et al. The metabolic landscape of lung cancer: new insights in a disturbed glucose metabolism. Front Oncol. 2019;9:1215. 10.3389/fonc.2019.01215. [DOI] [PMC free article] [PubMed] [Google Scholar]
44.Rauckhorst AJ, Sheldon RD, Pape DJ, et al. A hierarchical hepatic de Novo lipogenesis substrate supply network utilizing pyruvate, acetate, and ketones. Cell Metab Jan. 2025;7(1):255–e2736. 10.1016/j.cmet.2024.10.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
45.Sun Q, Huang M, Wei Y. Diversity of the reaction mechanisms of SAM-dependent enzymes. Acta Pharm Sin B. 2021;11(3):632–50. 10.1016/j.apsb.2020.08.011. Mar. [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Zheng L, Song Q, Xu X, et al. Structure-based insights into recognition and regulation of SAM-sensing riboswitches. Sci China Life Sci Jan. 2023;66(1):31–50. 10.1007/s11427-022-2188-7. [DOI] [PubMed] [Google Scholar]
47.Li Y, Xu T, Ma H, et al. Functional profiling of serine, threonine and tyrosine sites. Nat Chem Biol Apr. 2025;21(4):532–43. 10.1038/s41589-024-01731-0. [DOI] [PubMed] [Google Scholar]
48.Khalaji A, Yancheshmeh FB, Farham F, et al. Don’t eat me/eat me signals as a novel strategy in cancer immunotherapy. Heliyon Oct. 2023;9(10):e20507. 10.1016/j.heliyon.2023.e20507. [DOI] [PMC free article] [PubMed] [Google Scholar]
49.Locasale JW. Serine, Glycine and one-carbon units: cancer metabolism in full circle. Nat Rev Cancer Aug. 2013;13(8):572–83. 10.1038/nrc3557. [DOI] [PMC free article] [PubMed] [Google Scholar]
50.Sánchez-Castillo A, Heylen E, Hounjet J, et al. Targeting serine/glycine metabolism improves radiotherapy response in non-small cell lung cancer. Br J Cancer Mar. 2024;130(4):568–84. 10.1038/s41416-023-02553-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
51.Zhang C, Yu JJ, Yang C, et al. Wild-type IDH1 maintains NSCLC stemness and chemoresistance through activation of the Serine biosynthetic pathway. Sci Transl Med Dec. 2023;13(726):eade4113. 10.1126/scitranslmed.ade4113. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material 1^{(1.3MB, docx)}

Data Availability Statement

Data can obtain with the permission of the corresponding author.

[CR1] 1.Yin X, Li Y, Wang H, et al. Small cell lung cancer transformation: from pathogenesis to treatment. Semin Cancer Biol Nov. 2022;86(Pt 2):595–606. 10.1016/j.semcancer.2022.03.006. [DOI] [PubMed] [Google Scholar]

[CR2] 2.Kim SY, Park HS, Chiang AC. Small cell lung cancer: A review. Jama Jun. 2025;3(21):1906–17. 10.1001/jama.2025.0560. [DOI] [PubMed] [Google Scholar]

[CR3] 3.Herzog BH, Devarakonda S, Govindan R. Overcoming chemotherapy resistance in SCLC. J Thorac Oncol Dec. 2021;16(12):2002–15. 10.1016/j.jtho.2021.07.018. [DOI] [PubMed] [Google Scholar]

[CR4] 4.Fischer B, Marinov M, Arcaro A. Targeting receptor tyrosine kinase signalling in small cell lung cancer (SCLC): what have we learned so far? Cancer Treat Rev Jun. 2007;33(4):391–406. 10.1016/j.ctrv.2007.01.006. [DOI] [PubMed] [Google Scholar]

[CR5] 5.Gomez-Randulfe I, Leporati R, Gupta B, Liu S, Califano R. Recent advances and future strategies in first-line treatment of ES-SCLC. Eur J Cancer Mar. 2024;200:113581. 10.1016/j.ejca.2024.113581. [DOI] [PubMed] [Google Scholar]

[CR6] 6.Cheng Y, Fan Y, Zhao Y, et al. Tislelizumab plus platinum and Etoposide versus placebo plus platinum and Etoposide as First-Line treatment for Extensive-Stage SCLC (RATIONALE-312): A Multicenter, Double-Blind, placebo-Controlled, Randomized, phase 3 clinical trial. J Thorac Oncol Jul. 2024;19(7):1073–85. 10.1016/j.jtho.2024.03.008. [DOI] [PubMed] [Google Scholar]

[CR7] 7.Chen D, Zou B, Li B, et al. Adebrelimab plus chemotherapy and sequential thoracic radiotherapy as first-line therapy for extensive-stage small-cell lung cancer (ES-SCLC): a phase II trial. EClinicalMedicine Sep. 2024;75:102795. 10.1016/j.eclinm.2024.102795. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR8] 8.Xie M, Vuko M, Rodriguez-Canales J, et al. Molecular classification and biomarkers of outcome with immunotherapy in extensive-stage small-cell lung cancer: analyses of the CASPIAN phase 3 study. Mol Cancer May. 2024;30(1):115. 10.1186/s12943-024-02014-x. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR9] 9.Yi M, Niu M, Xu L, Luo S, Wu K. Regulation of PD-L1 expression in the tumor microenvironment. J Hematol Oncol Jan. 2021;7(1):10. 10.1186/s13045-020-01027-5. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR10] 10.Nabet BY, Hamidi H, Lee MC, et al. Immune heterogeneity in small-cell lung cancer and vulnerability to immune checkpoint Blockade. Cancer Cell Mar. 2024;11(3):429–e4434. 10.1016/j.ccell.2024.01.010. [DOI] [PubMed] [Google Scholar]

[CR11] 11.Paz-Ares L, Garassino MC, Chen Y, et al. Durvalumab ± Tremelimumab + Platinum-Etoposide in Extensive-Stage small cell lung cancer (CASPIAN): outcomes by PD-L1 expression and tissue tumor mutational burden. Clin Cancer Res Feb. 2024;16(4):824–35. 10.1158/1078-0432.Ccr-23-1689. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR12] 12.Xue Y, Gao S, Gou J, et al. Platinum-based chemotherapy in combination with PD-1/PD-L1 inhibitors: preclinical and clinical studies and mechanism of action. Expert Opin Drug Deliv Feb. 2021;18(2):187–203. 10.1080/17425247.2021.1825376. [DOI] [PubMed] [Google Scholar]

[CR13] 13.Chen T, Wang M, Chen Y, Cao Y, Liu Y. Advances in predictive biomarkers associated with immunotherapy in extensive-stage small cell lung cancer. Cell Biosci Sep. 2024;12(1):117. 10.1186/s13578-024-01283-9. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR14] 14.Paz-Ares L, Dvorkin M, Chen Y, et al. Durvalumab plus platinum-etoposide versus platinum-etoposide in first-line treatment of extensive-stage small-cell lung cancer (CASPIAN): a randomised, controlled, open-label, phase 3 trial. Lancet Nov. 2019;23(10212):1929–39. 10.1016/s0140-6736(19)32222-6. [DOI] [PubMed] [Google Scholar]

[CR15] 15.Liu SV, Reck M, Mansfield AS, et al. Updated overall survival and PD-L1 subgroup analysis of patients with Extensive-Stage Small-Cell lung cancer treated with Atezolizumab, Carboplatin, and Etoposide (IMpower133). J Clin Oncol Feb. 2021;20(6):619–30. 10.1200/jco.20.01055. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR16] 16.Killock D. Prospective insights on the use of bTMB. Nat Rev Clin Oncol Jun. 2022;19(6):360. 10.1038/s41571-022-00640-2. [DOI] [PubMed] [Google Scholar]

[CR17] 17.Xu Z, Zeng S, Gong Z, Yan Y. Exosome-based immunotherapy: a promising approach for cancer treatment. Mol Cancer Nov. 2020;12(1):160. 10.1186/s12943-020-01278-3. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR18] 18.Stadler JC, Belloum Y, Deitert B, et al. Current and future clinical applications of ctdna in Immuno-Oncology. Cancer Res Feb. 2022;1(3):349–58. 10.1158/0008-5472.Can-21-1718. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR19] 19.Yi K, Wang X, Filippov SK, Zhang H. Emerging ctdna detection strategies in clinical cancer theranostics. Smart Med Nov. 2023;2(4):e20230031. 10.1002/smmd.20230031. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR20] 20.Huo Z, Duan Y, Zhan D, et al. Proteomic stratification of prognosis and treatment options for small cell lung cancer. Genom Proteom Bioinf. Jul 2024;3(2). 10.1093/gpbjnl/qzae033. [DOI] [PMC free article] [PubMed]

[CR21] 21.Solta A, Ernhofer B, Boettiger K, et al. Small cells - big issues: biological implications and preclinical advancements in small cell lung cancer. Mol Cancer Feb. 2024;24(1):41. 10.1186/s12943-024-01953-9. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR22] 22.Chen H, Deng C, Gao J, et al. Integrative Spatial analysis reveals tumor heterogeneity and immune colony niche related to clinical outcomes in small cell lung cancer. Cancer Cell Mar. 2025;10(3):519–e5365. 10.1016/j.ccell.2025.01.012. [DOI] [PubMed] [Google Scholar]

[CR23] 23.Liu Q, Zhang J, Guo C, et al. Proteogenomic characterization of small cell lung cancer identifies biological insights and subtype-specific therapeutic strategies. Cell Jan. 2024;4(1):184–e20328. 10.1016/j.cell.2023.12.004. [DOI] [PubMed] [Google Scholar]

[CR24] 24.Cheng Y, Chen J, Zhang W, et al. Benmelstobart, anlotinib and chemotherapy in extensive-stage small-cell lung cancer: a randomized phase 3 trial. Nat Med Oct. 2024;30(10):2967–76. 10.1038/s41591-024-03132-1. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR25] 25.Shen G, Zheng F, Ren D, et al. Anlotinib: a novel multi-targeting tyrosine kinase inhibitor in clinical development. J Hematol Oncol Sep. 2018;19(1):120. 10.1186/s13045-018-0664-7. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR26] 26.Han X, Guo J, Li L, et al. Sintilimab combined with anlotinib and chemotherapy as second-line or later therapy in extensive-stage small cell lung cancer: a phase II clinical trial. Signal Transduct Target Ther Sep. 2024;16(1):241. 10.1038/s41392-024-01957-3. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR27] 27.Sen T, Takahashi N, Chakraborty S, et al. Emerging advances in defining the molecular and therapeutic landscape of small-cell lung cancer. Nat Rev Clin Oncol Aug. 2024;21(8):610–27. 10.1038/s41571-024-00914-x. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR28] 28.Thai AA, Solomon BJ, Sequist LV, Gainor JF, Heist RS. Lung cancer. Lancet. Aug 7. 2021;398(10299):535–554. 10.1016/s0140-6736(21)00312-3 [DOI] [PubMed]

[CR29] 29.Hendriks LE, Kerr KM, Menis J, et al. Oncogene-addicted metastatic non-small-cell lung cancer: ESMO clinical practice guideline for diagnosis, treatment and follow-up. Ann Oncol Apr. 2023;34(4):339–57. 10.1016/j.annonc.2022.12.009. [DOI] [PubMed] [Google Scholar]

[CR30] 30.Bauermeister A, Mannochio-Russo H, Costa-Lotufo LV, Jarmusch AK, Dorrestein PC. Mass spectrometry-based metabolomics in Microbiome investigations. Nat Rev Microbiol Mar. 2022;20(3):143–60. 10.1038/s41579-021-00621-9. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR31] 31.Wu Z, Bagarolo GI, Thoröe-Boveleth S, Jankowski J. Lipidomics: mass spectrometric and chemometric analyses of lipids. Adv Drug Deliv Rev. 2020;159:294–307. 10.1016/j.addr.2020.06.009. [DOI] [PubMed] [Google Scholar]

[CR32] 32.Wang R, Li B, Lam SM, Shui G. Integration of lipidomics and metabolomics for in-depth Understanding of cellular mechanism and disease progression. J Genet Genomics Feb. 2020;20(2):69–83. 10.1016/j.jgg.2019.11.009. [DOI] [PubMed] [Google Scholar]

[CR33] 33.Yu J, Xu H, Xiong F, et al. Lipidomics reveals biomarkers of the efficacy of first-line ICI therapy combined with chemotherapy in NSCLC. J Transl Med Jun. 2025;10(1):638. 10.1186/s12967-025-06542-y. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR34] 34.Xiong J, Xia L. Efficacy and safety of anlotinib as maintenance treatment in extensive-stage small cell lung cancer: a single-armed single center retrospective study. Front Oncol. 2024;14:1462581. 10.3389/fonc.2024.1462581. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR35] 35.Han B, Li K, Wang Q, et al. Effect of anlotinib as a Third-Line or further treatment on overall survival of patients with advanced Non-Small cell lung cancer: the ALTER 0303 phase 3 randomized clinical trial. JAMA Oncol Nov. 2018;1(11):1569–75. 10.1001/jamaoncol.2018.3039. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR36] 36.Xue MY, Sun HZ, Wu XH, Liu JX, Guan LL. Multi-omics reveals that the rumen microbiome and its metabolome together with the host metabolome contribute to individualized dairy cow performance. Microbiome. May 12. 2020;8(1):64. 10.1186/s40168-020-00819-8 [DOI] [PMC free article] [PubMed]

[CR37] 37.Rossi M, Altea-Manzano P, Demicco M, et al. PHGDH heterogeneity potentiates cancer cell dissemination and metastasis. Nature May. 2022;605(7911):747–53. 10.1038/s41586-022-04758-2. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR38] 38.Zhang D, Li AM, Hu G, et al. PHGDH-mediated endothelial metabolism drives glioblastoma resistance to chimeric antigen receptor T cell immunotherapy. Cell Metab Mar. 2023;7(3):517–e5348. 10.1016/j.cmet.2023.01.010. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR39] 39.Li SR, Bu LL, Cai L. Cuproptosis: lipoylated TCA cycle proteins-mediated novel cell death pathway. Signal Transduct Target Ther May. 2022;13(1):158. 10.1038/s41392-022-01014-x. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR40] 40.Zhan D, Wang D, Min W, Han W. Exploring the molecular basis for selective binding of Homoserine dehydrogenase from Mycobacterium leprae TN toward inhibitors: a virtual screening study. Int J Mol Sci Jan. 2014;24(2):1826–41. 10.3390/ijms15021826. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR41] 41.Mitchell SC, Waring RH. The deficiency of sulfoxidation of S-carboxymethyl-L-cysteine. Pharmacol Ther. 1989;43(2):237–49. 10.1016/0163-7258(89)90120-4. [DOI] [PubMed] [Google Scholar]

[CR42] 42.Pokorný V, Štejfa V, Havlín J, Fulem M, Růžička K. Heat capacities of L-Cysteine, L-Serine, L-Threonine, L-Lysine, and L-Methionine. Molecules Jan. 2023;3(1). 10.3390/molecules28010451. [DOI] [PMC free article] [PubMed]

[CR43] 43.Vanhove K, Graulus GJ, Mesotten L, et al. The metabolic landscape of lung cancer: new insights in a disturbed glucose metabolism. Front Oncol. 2019;9:1215. 10.3389/fonc.2019.01215. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR44] 44.Rauckhorst AJ, Sheldon RD, Pape DJ, et al. A hierarchical hepatic de Novo lipogenesis substrate supply network utilizing pyruvate, acetate, and ketones. Cell Metab Jan. 2025;7(1):255–e2736. 10.1016/j.cmet.2024.10.013. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR45] 45.Sun Q, Huang M, Wei Y. Diversity of the reaction mechanisms of SAM-dependent enzymes. Acta Pharm Sin B. 2021;11(3):632–50. 10.1016/j.apsb.2020.08.011. Mar. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR46] 46.Zheng L, Song Q, Xu X, et al. Structure-based insights into recognition and regulation of SAM-sensing riboswitches. Sci China Life Sci Jan. 2023;66(1):31–50. 10.1007/s11427-022-2188-7. [DOI] [PubMed] [Google Scholar]

[CR47] 47.Li Y, Xu T, Ma H, et al. Functional profiling of serine, threonine and tyrosine sites. Nat Chem Biol Apr. 2025;21(4):532–43. 10.1038/s41589-024-01731-0. [DOI] [PubMed] [Google Scholar]

[CR48] 48.Khalaji A, Yancheshmeh FB, Farham F, et al. Don’t eat me/eat me signals as a novel strategy in cancer immunotherapy. Heliyon Oct. 2023;9(10):e20507. 10.1016/j.heliyon.2023.e20507. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR49] 49.Locasale JW. Serine, Glycine and one-carbon units: cancer metabolism in full circle. Nat Rev Cancer Aug. 2013;13(8):572–83. 10.1038/nrc3557. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR50] 50.Sánchez-Castillo A, Heylen E, Hounjet J, et al. Targeting serine/glycine metabolism improves radiotherapy response in non-small cell lung cancer. Br J Cancer Mar. 2024;130(4):568–84. 10.1038/s41416-023-02553-y. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR51] 51.Zhang C, Yu JJ, Yang C, et al. Wild-type IDH1 maintains NSCLC stemness and chemoresistance through activation of the Serine biosynthetic pathway. Sci Transl Med Dec. 2023;13(726):eade4113. 10.1126/scitranslmed.ade4113. [DOI] [PubMed] [Google Scholar]

PERMALINK

Multi-omics dynamic profiling reveals predictive biomarkers for first-line immunochemotherapy in extensive-stage small-cell lung cancer

Liang Zheng

Haoming Xu

Shuyuan Wang

Meili Ma

Fang Hu

Lei Cheng

Jun Lu

Feng Pan

Bo Zhang

Jianlin Xu

Ying Li

Yinchen Shen

Wei Zhang

Runbo Zhong

Tianqing Chu

Baohui Han

Xiaoxuan Zheng

Hua Zhong

Wei Nie

Xueyan Zhang

Abstract

Background

Methods

Results

Conclusion

Supplementary Information

Introduction

Methods

Patient selection

Patient treatment and specimen collection

Untargeted metabolomics

Broadly targeted lipidomics

Ultra-fast hematoproteomics

Study design

Data analysis and statistical methods

Results

Baseline characteristics of patients

Fig. 1.

Prognosis of patients receiving different treatments

Fig. 2.

Machine learning approaches to screen differential metabolites

Fig. 3.

Machine learning approaches for lipid screening

Fig. 4.

Construction of a prognostic model

Fig. 5.

Proteomics analysis and integrated multi-omics analysis

Fig. 6.

Serum multi-omics profile after one cycle of treatment

Fig. 7.

Discussion

Supplementary Information

Acknowledgements

Author contributions

Funding

Data availability

Declarations

Ethical approval

Consent for publication

Generative AI and AI-assisted technologies

Competing interests

Footnotes

Contributor Information

References

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases