Abstract

   Bronchial asthma is a complex and heterogeneous disease, with
   ferroptosis, a form of non-apoptotic cell death, contributing to its
   pathogenesis by inducing airway epithelial damage, inflammatory
   infiltration, and airway remodeling. Investigating ferroptosis-related
   characteristic genes and potential therapeutic compounds may enhance
   asthma management. This study employed differential analysis and
   machine learning to identify ferroptosis-related characteristic genes
   in asthma using the [40]GSE179156 dataset and FerrDb V2 database.
   Immune infiltration analysis explored the associations between these
   genes and immune cells, while potential small-molecule drugs were
   screened through the Connectivity Map (CMap) database and evaluated via
   molecular docking and molecular dynamics simulations. Two
   ferroptosis-related characteristic genes, AGPS and APELA, were
   identified, with AGPS upregulated and APELA downregulated in asthma,
   both significantly correlated with various immune cells. A diagnostic
   model based on these genes demonstrated high predictive accuracy.
   Additionally, KU-55933 was identified as a potential small-molecule
   inhibitor of AGPS, with stable binding confirmed through computational
   simulations. These findings emphasize the role of ferroptosis-related
   genes in asthma and propose promising therapeutic candidates, providing
   novel insights into its diagnosis and treatment.

   Keywords: Asthma, Ferroptosis, Machine learning, Hub genes, Potential
   therapeutic compounds

   Subject terms: Computational biology and bioinformatics, Diseases

Introduction

   Bronchial asthma is a heterogeneous disease influenced by a combination
   of genetic and environmental factors. Clinically, it is characterized
   by recurrent episodes of wheezing, coughing, chest tightness, and
   shortness of breath due to airway obstruction and hyperreactivity, with
   chronic airway inflammation, excessive mucus production, and airway
   remodeling as pathological hallmarks^[41]1. Global data from 2019
   report an age-standardized prevalence of approximately 3415.5 per
   100,000 population and an all-cause mortality rate of about 5.8 per
   100,000^[42]2. Asthma often coexists with conditions such as allergic
   rhinitis, obesity, and attention-deficit hyperactivity disorder in
   children^[43]3–[44]6, Patients with asthma also face a higher risk of
   all-cause mortality^[45]7,[46]8, leading to a significant reduction in
   quality of life and imposing substantial economic burdens. The complex
   mechanisms underlying asthma, coupled with its diverse phenotypes and
   endotypes, pose significant challenges for precise diagnosis and
   effective treatment^[47]9,[48]10. Identifying diagnostic biomarkers for
   asthma could improve early detection and facilitate better disease
   management.

   Recent studies have demonstrated that various forms of cell death,
   including autophagy, necroptosis, pyroptosis, and ferroptosis, play
   roles in asthma pathogenesis^[49]11. Among these, ferroptosis—a
   non-apoptotic form of cell death driven by imbalances in iron
   metabolism and lipid peroxidation—has been shown to play a critical
   role in lung diseases^[50]12,[51]13. Ferroptosis exacerbates airway
   inflammation, induces epithelial damage, and promotes airway remodeling
   in asthma. These processes involve the regulation of proteins such as
   PEBP1, 15LO-1, GPX-4, and SLC7A11. Interestingly, ferroptosis
   inhibitors or interventions targeting these proteins have been found to
   alleviate airway inflammation and epithelial damage^[52]14–[53]16.
   These findings highlight the importance of ferroptosis in asthma
   pathogenesis and suggest its potential as a novel therapeutic target.

   KU-55933, a well-known non-competitive inhibitor of Ataxia
   telangiectasia mutated (ATM) kinase, has been widely used to counteract
   DNA repair mechanisms in tumor cells, thereby enhancing chemotherapy
   sensitivity^[54]17. Recent evidence suggests that ATM kinase promotes
   ferroptosis by activating NCOA4, while KU-55933 inhibits ATM activation
   and suppresses ferroptosis mediated by the 15LO2-PEBP1
   pathway^[55]18,[56]19. These findings indicate that KU-55933 may have
   therapeutic potential beyond its conventional applications, making it a
   promising candidate for further exploration.

   This study aims to investigate the role of ferroptosis in asthma
   pathogenesis and identify potential diagnostic biomarkers.
   Additionally, it explores the preliminary therapeutic potential of
   small-molecule compounds KU-55933, as innovative approaches for asthma
   treatment.

Methods

Data collection and processing

   Asthma-related transcriptome datasets were screened from the GEO
   database ([57]https://www.ncbi.nlm.nih.gov/geo/). Inclusion criteria
   were as follows: datasets based on array expression, samples from lower
   airway tissues, studies involving humans, datasets including both
   asthma and control groups with at least five samples per group, and
   availability of transcriptome data. Following these criteria,
   [58]GSE179156^[59]20 was included as the training dataset, consisting
   of airway epithelial cell transcriptomes from 57 asthma patients and 29
   healthy controls. Additionally, [60]GSE41861 and [61]GSE104468^[62]21
   were included for analysis. [63]GSE41861 contained nasal and airway
   epithelial cell transcriptomes from 51 asthma patients and 30 healthy
   controls; however, only lower airway data were included due to its
   relevance to asthma as the primary site of pathology. Similarly, from
   [64]GSE104468, which included nasal, airway epithelial, and Peripheral
   blood mononuclear cell (PBMC) transcriptomes of 12 asthma patients and
   12 healthy controls, only the lower airway data were used. The original
   studies for all three datasets were independently conducted by research
   institutions in the United States, with enrolled participants being
   adult subjects. In our study, data from different sources were analyzed
   separately as training or validation sets, which helps mitigate
   potential heterogeneity between datasets of different origins.
   Furthermore, all selected data in this study were derived from lower
   airway tissue transcriptomes, thereby avoiding gene expression
   variations across different tissue types. Ferroptosis-related genes
   (FRGs) were downloaded from the FerrDb V2 database
   ([65]http://www.zhounan.org/ferrdb/current/)^[66]22 on October 13,
   2024. FerrDb V2 contains 1,001 regulators and 143 diseases related to
   ferroptosis, from which 525 FRGs were identified (Supplementary file
   1). Detailed information on the datasets is presented in Table [67]1.
   The flowchart of this study is shown in Fig. [68]1.

Table 1.

   Detailed information on datasets used in the study.
   Dataset/Database Platforms Sample size Organism Tissue Attribute
   Asthma/Ferroptosis Control
   [69]GSE179156 [70]GPL570 57 29 Homo Sapiens Bronchial Epithelia
   Training
   [71]GSE41861 [72]GPL570 51 30 Homo Sapiens Bronchial Epithelia
   Validation
   [73]GSE104468 [74]GPL21185 12 12 Homo Sapiens Bronchial Epithelia
   Validation
   Ferroptosis related genes FerrDb V2 525 – Homo Sapiens – –
   [75]Open in a new tab

Fig. 1.

   [76]Fig. 1
   [77]Open in a new tab

   The flowchart of this study. Sequence of all workflows in this study.

Identification of differentially expressed genes (DEGs)

   Data analysis was performed using R 4.2.1, and differential gene
   expression was identified with the “limma” package. Screening criteria
   were set to adj.P < 0.05 and |Log2FC|> 0.58. Volcano plots and heatmaps
   were generated using the “ggplot2” and “pheatmap” packages,
   respectively. Intersection genes between DEGs and FRGs were identified
   using the “VennDiagram” package, defined as Ferr-DEGs, and used for
   subsequent analyses.

Correlation analysis and PPI network construction

   The “Rcircos” package was used to generate landscape maps of Ferr-DEGs.
   A protein–protein interaction (PPI) network was constructed using the
   STRING database ([78]https://cn.string-db.org/)^[79]23.

GO and KEGG pathway enrichment analyses

   Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG)
   pathway enrichment analyses of Ferr-DEGs were performed using the
   enrichGO and enrichKEGG functions in the “clusterProfiler” package,
   with the screening criteria set to pvalueCutoff = 0.05 and
   qvalueCutoff = 0.05 for both analyses. Additionally, single-gene Gene
   Set Enrichment Analysis (GSEA) was conducted to assess the correlations
   between diagnostic genes and other genes, followed by further
   enrichment analysis using the gseGO and gseKEGG functions.

Machine learning

   Three machine learning algorithms—Random Forest (RF), Lasso regression,
   and Boruta—were applied to identify core genes. Lasso regression
   selects key predictors by shrinking coefficients of less relevant
   variables to zero, mitigating overfitting risks. In this study, a
   binomial family was specified, with the alpha penalty parameter set to
   1 and nlambda configured to 1000 for Lasso regression. The RF algorithm
   combines predictions from multiple decision trees trained on random
   data subsets to enhance accuracy and robustness. During the RF-based
   feature selection process, the response variable was set as the group
   data, while the predictor variables consisted of gene expression data.
   The number of trees (ntree) was set to 500, after which the model with
   the lowest prediction error was used to construct the optimal random
   forest model for gene selection. Boruta, an extension of random forest,
   iteratively compares original features with shadow features to identify
   significant variables. The “glmnet”, “randomForest”, and “Boruta”
   packages were used for respective analyses, and the intersection of
   results from the three algorithms was considered the final core gene
   set for further analysis.

Diagnostic model construction

   The “rms” package was used to construct a nomogram to evaluate the
   diagnostic performance of core genes. Calibration curves assessed the
   accuracy of the nomogram, which represented both individual and
   combined predictive capabilities of the genes. A Decision Curve
   Analysis (DCA) evaluated the diagnostic utility of the model. External
   validation was performed using [80]GSE41861 and [81]GSE104468 datasets.

Immune infiltration analysis

   The “CIBERSORT”^[82]24 package was used to calculate the proportions of
   various immune cells in each sample. Results were visualized using the
   “corrplot” package. Spearman correlation analysis was conducted to
   examine associations between core genes and immune scores.

Compound screening using the CMap database

   The Connectivity Map (CMap, [83]http://clue.io/)^[84]25 database was
   used to identify small-molecule compounds potentially related to asthma
   treatment. Upregulated Ferr-DEGs from asthma datasets were uploaded to
   the CMap database, and compounds were ranked by enrichment scores in
   ascending order. The top 10 compounds were considered as potential
   therapeutic candidates.

Molecular docking and molecular dynamics simulations

   AGPS sequence information was obtained from the UniProt
   ([85]https://www.uniprot.org/) database and its 3D structure predicted
   using the AlphaFold3 (AF3) AI program
   ([86]https://alphafoldserver.com/)^[87]26. Molecular structures of
   small compounds were retrieved from the PubChem
   ([88]https://pubchem.ncbi.nlm.nih.gov/) database. AutoDock 1.5.7
   software was used for molecular docking, and results were visualized
   using PyMOL 2.6.0 software. Lower binding energies indicated stronger
   protein–ligand interactions. The docked complexes were imported into
   Gromacs 2.0 software for molecular dynamics simulations, and
   protein–ligand binding free energies were calculated using the
   Molecular Mechanics/Poisson-Boltzmann Surface Area (MM/PBSA) method.

Results

Identification of DEGs in asthma

   Principal Component Analysis (PCA) revealed distinct differences in
   gene expression between asthma and control groups in the [89]GSE179156
   dataset (Fig. [90]2a). A total of 758 DEGs were identified using the
   Limma package, including 361 upregulated and 397 downregulated genes in
   asthma samples (P < 0.05, |Log2FC|≥ 0.58). A volcano plot summarizing
   the DEGs was generated and is shown in Fig. [91]2b. Hierarchical
   clustering analysis was performed, and the results were visualized with
   a heatmap (Fig. [92]2c).

Fig. 2.

   [93]Fig. 2
   [94]Open in a new tab

   Screening differentially expressed genes (DEGs) in asthma. (a)
   Principal component analysis (PCA) illustrating distinct differences in
   gene expression between asthma and control groups in the GSE dataset.
   (b) Volcano plot highlighting upregulated and downregulated genes
   (n = 758, P < 0.05). (c) Heatmap of hierarchical clustering analysis
   showing the expression patterns of differentially expressed genes
   (DEGs) between asthma and control samples in [95]GSE179156. (d) Gene
   Ontology (GO) enrichment analysis of DEGs in GSE 179,156, encompassing
   biological processes, cellular components, and molecular functions. (e)
   KEGG pathway enrichment analysis of DEGs in GSE 179,156, demonstrating
   the involvement of key signaling pathways.

   GO and KEGG enrichment analyses were subsequently conducted. The GO
   analysis (Fig. [96]2d) indicated that DEGs were involved in biological
   processes (BP) such as taxis, chemotaxis, and leukocyte migration. For
   cellular components (CC), DEGs were enriched in the collagen-containing
   extracellular matrix, basal part of the cell, and basal plasma
   membrane. Molecular functions (MF) included enzyme inhibitor activity,
   peptidase regulator activity, and endopeptidase regulator activity.
   KEGG pathway analysis (Fig. [97]2e) demonstrated that DEGs were
   significantly enriched in pathways such as cytokine-cytokine receptor
   interaction, IL-17 signaling pathway, amoebiasis, pertussis, and TNF
   signaling pathway.

   These enrichment analyses suggested that asthma pathogenesis involved
   intricate mechanisms, such as immune responses, inflammatory processes,
   and interactions between cytokines and their receptors, highlighting
   the complex systems underlying the disease.

Ferr-DEGs and correlation analysis

   Intersecting the 758 asthma DEGs with 525 ferroptosis-related genes
   resulted in 21 overlapping genes, termed Ferr-DEGs (Fig. [98]3a).
   Spearman correlation analysis revealed associations among Ferr-DEGs
   (Fig. [99]3b). Chromosomal localization of these genes was mapped
   (Fig. [100]3c). GO enrichment analysis (Fig. [101]3d) showed that
   Ferr-DEGs were significantly enriched in biological processes,
   including ameboid-like cell migration, epithelial cell migration,
   smooth muscle proliferation, and cellular responses to metal ions,
   suggesting their potential roles in cellular motility and stress
   responses. KEGG analysis (Fig. [102]3e) indicated enrichment in
   pathways related to fluid shear stress and atherosclerosis, IL-17
   signaling, lipid metabolism, AGE-RAGE signaling in diabetes, and Th17
   differentiation. STRING database analysis demonstrated potential PPI
   among these genes (Fig. [103]3f).

Fig. 3.

   [104]Fig. 3
   [105]Open in a new tab

   Screening ferroptosis related differentially expressed genes
   (Ferr-DEGs) in asthma. (a) Venn diagram illustrating the overlap
   between asthma-related DEGs and ferroptosis-related genes (Ferr-DEGs).
   (b) Heatmap showing the Spearman correlation among Ferr-DEGs. (c)
   Chromosomal landscape map displaying the specific locations of
   Ferr-DEGs. (d) GO analysis includes biological process, cellular
   component, and molecular function of Ferr-DEGs. (e) KEGG analysis of
   Ferr-DEGs. (f) Protein–protein interaction (PPI) network of Ferr-DEGs.

   In asthma, ferroptosis-related pathways are closely associated with
   lipid peroxidation and iron ion accumulation. Elevated expression of
   molecules such as DMT1 and TFR1 in airway tissues promotes iron
   accumulation in epithelial cells, which triggers smooth muscle
   proliferation, enhances pro-inflammatory fibroblast activity, and
   stimulates excessive extracellular matrix production, ultimately
   driving airway remodeling^[106]27. Additionally, chronic airway
   inflammation further activates ferroptosis-related pathways, including
   the 15LO1-PEBP1 axis, exacerbating airway remodeling and sustaining
   inflammatory responses^[107]14. These findings highlight the critical
   role of ferroptosis in asthma pathogenesis, suggesting that
   investigating Ferr-DEGs could provide insights into the underlying
   molecular mechanisms and pave the way for the development of targeted
   therapeutic strategies.

Identification of asthma-specific genes

   Lasso (Fig. [108]4b) identified 13 genes, including APELA, NOX1, EMP1,
   GRIA3, AEBP1, NR4A1, APOE, AGPS, TRPM2, CD44, NFE2L2, MAPK8, and DPP4
   (λ1se = 0.06860). RF (Fig. [109]4d) ranked genes by importance and
   selected 10 genes with importance > 1.5, including APELA, MAPK8, AGPS,
   AEBP1, MT1G, GJA1, NOX1, IDO1, IL6, and APOC1. Boruta (Fig. [110]4e)
   identified 11 genes: APELA, MT1G, NOX1, GJA1, EMP1, AEBP1, NR4A1, AGPS,
   TRPM2, CD44, and MAPK8 ranked by significance. The intersection of the
   three methods yielded five genes: APELA, NOX1, AEBP1, AGPS, and MAPK8
   (Fig. [111]4g). Multivariable logistic regression (Fig. [112]4h)
   confirmed that AGPS (OR 4.4099, P < 0.05) and APELA (OR 0.3359,
   P < 0.05) were significantly associated with asthma, establishing them
   as characteristic asthma genes.

Fig. 4.

   [113]Fig. 4
   [114]Open in a new tab

   Identification of key genes in asthma using machine learning and
   Feature gene selection. (a) Coefficient profiles of variables in the
   LASSO regression model. (b) Ten cross-validation for turning parameter
   (λ) selection in the LASSO regression model. (c) The error rate curve
   of the Random Forest model as a function of the number of trees. (d) 10
   genes with importance selected by Random Forest model. (e) Variable
   importance ranking in the Boruta algorithm. (f) The variation trend of
   Z-scores in the Boruta algorithm. (g) Venn diagram of RF, LASSO and
   BORUTA. (h) Multivariable logistic regression confirmed that AGPS and
   APELA were significantly associated with asthma.

GSEA analysis of characteristic genes

   GSEA analysis of APELA and AGPS revealed their involvement in
   biological processes such as vascular and circulatory system
   development, immune regulation, and molecular function modulation.
   Pathways such as cytokine-cytokine receptor interactions and RAS
   signaling were also identified as significant. These findings suggested
   that these genes might play critical roles in cell development and the
   regulation of immune infiltration (Fig. [115]5a–c).

Fig. 5.

   [116]Fig. 5
   [117]Open in a new tab

   Single gene GSEA of characteristic genes. (a) GO results of GSEA
   enrichment analysis of AGPS. (b) KEGG results of GSEA enrichment
   analysis of AGPS. (c) GO results of GSEA enrichment analysis of APELA.

   Notably, APELA was not included in the KEGG subset of the GSEA
   database. As a result, only GO enrichment analysis was performed for
   APELA, and KEGG enrichment analysis could not be conducted.

Diagnostic performance evaluation of characteristic genes

   A nomogram constructed using the rms package evaluated the diagnostic
   performance of APELA and AGPS, demonstrating strong predictive
   capabilities (Fig. [118]6a). Calibration curves indicated high
   consistency between predicted and observed outcomes (Fig. [119]6b).
   Decision curve analysis (DCA) further highlighted the clinical
   applicability of the diagnostic model (Fig. [120]6c). AGPS expression
   was significantly upregulated in asthma patients, whereas APELA
   expression was downregulated (Fig. [121]6d). Receiver operating
   characteristic (ROC) analysis revealed robust diagnostic efficiency for
   AGPS (AUC = 0.805) and APELA (AUC = 0.830) (Fig. [122]6e,f). The
   combined nomogram achieved an AUC of 0.880, indicating high
   discriminative power (Fig. [123]6g).

Fig. 6.

   [124]Fig. 6
   [125]Open in a new tab

   Construction of the diagnostic nomogram and diagnostic performance
   assessment. (a) The nomogram to predicted the diagnostic performance of
   APELA and AGPS. (b) The calibration curves of diagnostic performance.
   (c) Decision curve of diagnostic performance. (d) Expression of AGPS
   and APELA in [126]GSE179156. (e) ROC curves of AGPS in [127]GSE179156.
   (f) ROC curves of APELA in [128]GSE179156. (g) ROC curves of AGPS and
   APELA in [129]GSE179156.

External validation of the characteristic genes and nomogram

   External validation datasets ([130]GSE41861 and [131]GSE104468)
   confirmed the findings, with AGPS and APELA demonstrating consistent
   expression patterns (Fig. [132]7d and h). The area under the curve
   (AUC) values further validated the diagnostic reliability of the model.
   In the [133]GSE41861 dataset, AGPS achieved an AUC of 0.783, APELA
   reached 0.837, and the nomogram attained 0.859 (Fig. [134]7a–c).
   Similarly, in the [135]GSE104468 dataset, AGPS showed an AUC of 0.743,
   APELA exhibited 0.833, and the nomogram achieved 0.882, underscoring
   the robustness of the diagnostic model (Fig. [136]7e–g).

Fig. 7.

   [137]Fig. 7
   [138]Open in a new tab

   Evaluation of two characteristic genes in external datasets. (a) ROC
   curves of AGPS in [139]GSE41861. (b) ROC curves of APELA in
   [140]GSE41861. (c) ROC curves of AGPS and APELA in [141]GSE41861. (d)
   Expression of AGPS and APELA in [142]GSE41861. (e) ROC curves of AGPS
   in [143]GSE104468. (f) ROC curves of APELA in [144]GSE104468. (g) ROC
   curves of AGPS and APELA in [145]GSE104468. (h) Expression of AGPS and
   APELA in [146]GSE104468.

Immune infiltration analysis

   Given the established connection between ferroptosis, immune
   dysregulation, and inflammation in asthma, immune cell infiltration was
   analyzed using the CIBERSORT algorithm. Figures [147]8a and b
   illustrated the proportional distribution of 22 immune cell subtypes
   across individual samples. Compared to controls, asthma samples
   exhibited significantly elevated proportions of naïve B cells, plasma
   cells, resting memory CD4 + T cells, activated dendritic cells, and
   resting mast cells (P < 0.05) (Supplementary Fig. 1). Spearman
   correlation analysis further revealed distinct associations between
   characteristic genes (AGPS and APELA) and immune cell subsets
   (Fig. [148]8c). Specifically, AGPS displayed positive correlations with
   naïve B cells and resting mast cells, whereas APELA showed negative
   correlations with these cell types (Fig. [149]8d–g). These findings
   underscored a potential mechanistic link between ferroptosis-mediated
   processes, immune cell dynamics, and asthma pathogenesis.

Fig. 8.

   [150]Fig. 8
   [151]Open in a new tab

   Immune cell infiltration analysis. (a,b) Immune Infiltration Analysis
   of [152]GSE179156. (c) Spearman correlation between AGPS and APELA and
   22 immune cells. (d) The correlation of AGPS and mast cells resting.
   (e) The correlation of AGPS and naive B cells. (f) The correlation of
   APELA and mast cells resting. (g) The correlation of APELA and naive B
   cells.

Identification of potential therapeutic compounds using CMAP

   The data for upregulated Ferr-DEGs were uploaded to the CMap database.
   Since there were only nine upregulated genes among the Ferr-DEGs, the
   most significant upregulated DEG, CLCA1 (Log2FC = 3.75, P < 0.05) was
   included to the dataset for analysis. The top ten compounds negatively
   correlated with asthma were identified as potential therapeutic
   candidates: aripiprazole, BMS-927711, KU-55933, lomeguatrib, ORG-9768,
   oxprenolol, phenylbutyrate, riociguat, ZK-93426, and zonisamide. The 3D
   structures of these small molecules were retrieved from the PubChem
   database to facilitate further investigations (Supplementary file2).

Molecular docking and dynamics simulation of KU-55933 with AGPS

   Molecular docking was performed between the ten previously identified
   potential compounds and the AGPS macromolecule. Six compounds showed
   binding energies greater than 5 kJ/mol: aripiprazole (7.61 kcal/mol),
   BMS-927711 (6.18 kcal/mol), KU-55933 (6.52 kcal/mol), ORG-9768
   (5.67 kcal/mol), riociguat (5.33 kcal/mol), and zonisamide
   (5.5 kcal/mol), while the remaining four compounds exhibited binding
   energies below this threshold: lomeguatrib (4.82 kcal/mol), oxprenolol
   (3.41 kcal/mol), phenylbutyrate (4.26 kcal/mol), ZK-93426
   (4.84 kcal/mol). Further analysis revealed that, except for KU-55933,
   none of the other compounds formed hydrogen bonds with the
   macromolecular protein. Consequently, only AGPS and KU-55933 were
   selected for subsequent analyses.

   Molecular docking revealed strong binding affinity between KU-55933 and
   AGPS, with a binding energy of − 6.52 kcal/mol. Key interaction sites
   included hydrogen bonds formed with residues ARG-177 (2.11 Å), and
   GLU-285 (2.21 Å), highlighting the molecular stability of the complex
   (Fig. [153]9a). Molecular dynamics simulations confirmed the stability
   of the AGPS/KU-55933 complex, which achieved equilibrium after 90 ns
   with minimal structural fluctuations (RMSD ~ 13 Å) (Fig. [154]9b). Root
   mean square fluctuation (RMSF) values for AGPS residues remained mostly
   below 3 Å, further indicating a rigid and stable connection between
   KU-55933 and AGPS (Fig. [155]9c). Additional analyses of the radius of
   gyration (Rg) and solvent-accessible surface area (SASA) demonstrated a
   compact and stable interaction (Fig. [156]9d). On average, the complex
   maintained two hydrogen bonds (range: 0–4), reflecting strong hydrogen
   bonding interactions (Fig. [157]9e). Binding free energy calculations
   using the MM/PBSA method resulted in a value of -106.102 kJ/mol
   (− 25.359 kcal/mol), confirming the strong binding affinity and
   pharmacological potential of KU-55933. Key residues contributing to the
   interaction included PRO119, LEU506, and TYR505, underscoring the
   reliability of these findings and the therapeutic promise of KU-55933
   in targeting ferroptosis pathways in asthma (Fig. [158]9f).

Fig. 9.

   [159]Fig. 9
   [160]Open in a new tab

   Molecular docking analysis and molecule dynamics. (a) Ligand–protein
   complexes formed between the AGPS and KU-55933 with hydrogen bonds
   forming at ARG-117 2.11 Å, and GLN-285 2.21 Å. (b) RMSD: The extent of
   atomic positional variation over time in protein complexes. (c) RMSF:
   The degree of deviation of atoms in protein complexes from their mean
   positions. (d) SASA: Solvent Accessible Surface Area. (e) Hydrogen Bond
   Formation. (f) MM/PBSA: The binding/interaction energy between ligands
   and receptors.

   MC3805, a known AGPS inhibitor, had been reported to significantly
   reduce ether lipid levels in melanoma, breast cancer, and ovarian
   cancer cells^[161]28. However, previous studies^[162]28,[163]29 had
   employed AGPS (pdb_00005adz) conformations inconsistent with those
   predicted by the UniProt database and AlphaFold3 (AF3). To address this
   discrepancy, molecular docking of MC3805 was performed with the
   AF3-predicted AGPS structure (AF-[164]O00116-F1). The results
   demonstrated stable binding between MC3805 and AGPS, with a binding
   energy of − 5.65 kcal/mol (Supplementary Fig. 2–3), thereby validating
   the structural accuracy of the AF3-predicted AGPS conformation and
   supporting the reliability of molecular docking studies based on this
   model. Collectively, these findings underscored the therapeutic
   potential of AGPS-targeting inhibitors such as KU-55933 and MC3805,
   highlighting their potential as candidates for therapeutic development
   in asthma.

Discussion

   Asthma is a highly heterogeneous airway disease with variable clinical
   presentations, pathological features, and prognoses. Ferroptosis has
   been implicated in asthma pathogenesis and progression, influencing
   prognosis. This study aimed to explore ferroptosis-related therapeutic
   approaches for asthma. Through transcriptomics analysis, we identified
   two key genes, AGPS and APELA, and developed a diagnostic nomogram for
   asthma. Immune cell infiltration analysis revealed immune dysregulation
   in asthma and its potential association with these characteristic
   genes. Additionally, potential therapeutic small molecules were
   screened using the CMap database. Molecular docking and dynamics
   simulations.

   Due to the complexity of asthma pathogenesis and treatment, previous
   studies have identified asthma biomarkers and characteristic genes from
   various perspectives. For example, Yan^[165]30 identified five
   characteristic genes associated with asthma and rhinitis comorbidity
   using differential gene expression analysis, focusing on biological
   processes without evaluating their diagnostic potential or providing
   experimental evidence. Similarly, Wang^[166]31 identified ARK1C3 as a
   ferroptosis-related gene in asthma and investigated its regulatory role
   in ferroptosis in BEAS-2B cells. However, potential drugs targeting
   ARK1C3 were not explored, and its therapeutic relevance in asthma
   remains unclear. In this study, we used machine learning and logistic
   regression to identify two ferroptosis-related characteristic genes,
   AGPS and APELA, with distinct expression trends in asthma. Using
   ferroptosis-related genes from the FerrDb V2 database, our research is
   grounded in documented evidence, enhancing its reliability. Further
   analysis demonstrated that the combined diagnostic efficiency of these
   genes was superior to individual genes, underscoring their critical
   roles in asthma.

   AGPS is a lipid metabolism enzyme involved in the synthesis of ether
   lipids by converting acyl-glycerone-phosphate to
   alkyl-glycerone-phosphate. AGPS has been implicated in several
   diseases, including tumor progression, by regulating the MAPK pathway
   and factors like Twist, AP-1, Snail, and MMP-2, as well as lipid
   metabolism pathways such as LPA and LPA2. These processes promote
   epithelial-mesenchymal transition (EMT), cell cycle progression,
   invasion, and metastasis^[167]32,[168]33. In hepatocellular carcinoma,
   AGPS supports cell proliferation via metabolism linked to HDGF^[169]34.
   In contrast, AGPS appears to have protective effects in certain
   conditions, such as maintaining ether lipid homeostasis in brain
   tissues to mitigate oxidative stress in Alzheimer’s disease^[170]35.
   The role of AGPS in lipid metabolism also makes it a key player in
   ferroptosis, where its overexpression promotes cell death^[171]36. In
   asthma, ferroptosis exacerbates airway inflammation and epithelial cell
   damage^[172]11. In this study, AGPS was upregulated in asthma samples
   and correlated with immune cells such as CD4 + T cells and M0
   macrophages, suggesting its involvement in airway inflammation and
   immune dysregulation.

   APELA is a protein that acts through the apelin peptide jejunum
   receptor and has been shown to exert protective effects in various
   diseases, including atherosclerosis, myocardial infarction, heart
   failure, and pulmonary hypertension (PAH). In human lung tissues, APELA
   is primarily expressed in pulmonary artery endothelial cells. Studies
   have identified APELA as a vascular endothelial protective factor, with
   its downregulation in chronic obstructive pulmonary disease
   contributing to endothelial-mesenchymal transition and vascular
   remodeling, ultimately worsening COPD-associated PAH^[173]37,[174]38.
   Research by Yang et al. further demonstrated that exogenous APELA
   supplementation can counteract its downregulation in PAH, inhibiting
   cardiopulmonary remodeling and functional impairments, highlighting
   APELA’s therapeutic potential^[175]37.Abnormal vascular remodeling in
   lung tissue is also a pathological feature of asthma, exacerbating
   airway edema and obstruction^[176]39–[177]41. Targeting vascular injury
   and abnormal angiogenesis may represent a novel therapeutic approach
   for asthma^[178]42. Given the established association between vascular
   damage, angiogenesis, and asthma, as well as APELA’s protective role in
   vascular health, the potential involvement of APELA in asthma warrants
   further investigation. In this study, we observed decreased APELA
   expression in asthma patients, suggesting its possible role as a
   biomarker for the disease. Its downregulation may contribute to asthma
   development, further emphasizing the need to explore APELA’s
   therapeutic and diagnostic potential in asthma.

   Current research on asthma is transitioning from phenotyping to
   endotyping, with treatment strategies targeting specific molecular
   biological functions and pathophysiological mechanisms becoming pivotal
   in asthma management^[179]43. Clinically established asthma biomarkers
   and their corresponding monoclonal antibodies primarily
   include^[180]44: anti-IgE (Omalizumab) for Th2 inflammation,
   anti-IL4/IL13 (Dupilumab), anti-IL5 (mepolizumab), as well as
   alarmin-related biomarkers like anti-TSLP (Tezepelumab). Although IL17
   and IL23 have been identified as biomarkers for neutrophilic asthma,
   their monoclonal antibodies failed to demonstrate expected efficacy in
   clinical trials (Secukinumab, [181]NCT01478360), with some even
   exacerbating asthma symptoms (Risankizumab, [182]NCT02443298)^[183]45.
   Current evidence indicates that biomarker-targeted biologics are mainly
   effective for severe Th2-high asthma, while targeted therapies for
   other subtypes require further development. Given the highly
   heterogeneous and compensatory nature of asthma pathogenesis,
   elucidating non-Th2 signaling pathways will facilitate precision
   medicine. This study reveals that AGPS and APELA not only show
   diagnostic potential but are also involved in ferroptosis regulation,
   suggesting these targets may offer novel therapeutic approaches for
   severe asthma.

   KU-55933, a selective ATM kinase inhibitor, had been extensively
   studied in cancer research for its ability to disrupt DNA damage repair
   by inhibiting ATM activity. However, its applications outside oncology
   remained underexplored^[184]46. ATM kinase played a critical role in
   maintaining cellular homeostasis and regulating immune and inflammatory
   responses^[185]47. Additionally, ATM influenced various cell death
   processes, including apoptosis, necroptosis, ferroptosis, and
   autophagy, through pathways involving MFT1, PEX-5/p62, and p53^[186]48.
   KU-55933 had been shown to suppress ferroptosis by upregulating 15LO2
   expression, effectively mitigating sevoflurane-induced
   neurotoxicity^[187]19. In this study, KU-55933 was identified as a
   potential therapeutic small molecule for asthma through the CMap
   database and machine learning-based predictions. Molecular docking and
   MD simulations demonstrated that KU-55933 stably bound to the active
   pocket of AGPS. To further evaluate the binding capacity of KU-55933 to
   AGPS, molecular docking of MC-3805 with AGPS was performed. The results
   revealed that KU-55933 exhibited a higher binding energy
   (− 6.52 kcal/mol) compared to MC-3805 (− 5.65 kcal/mol) and formed two
   stable hydrogen bonds with AGPS (Supplementary Fig. 2–3), confirming
   the superior binding capacity of KU-55933 to AGPS. Subsequent molecular
   dynamics simulations corroborated these findings, demonstrating stable
   interactions between KU-55933 and AGPS. These findings suggested that
   KU-55933 might function as a small-molecule inhibitor of AGPS,
   potentially counteracting ferroptosis-induced airway epithelial damage
   in asthma. Based on the biological characteristics of AGPS, we
   speculate that KU-55933 may exert therapeutic effects by inhibiting
   AGPS activity, which would consequently suppress multiple signaling
   pathways including MEK/ERK/SRF^[188]49. This inhibition could lead to
   downregulation of various signaling molecules such as E-cadherin,
   Snail, and MMP2^[189]33. Simultaneously, it may inhibit ferroptosis in
   pulmonary tissue cells, alleviate airway EMT in asthma, and reduce
   airway inflammation. Further research is warranted to validate its
   therapeutic potential in clinical settings and to examine the
   differences in effects compared to MC3805, in order to clarify the
   potential of KU-55933 for asthma treatment.

   Despite the promising findings, this study has several limitations.
   First, while we identified two ferroptosis-related characteristic genes
   through computational analyses and validated their diagnostic
   capabilities with external datasets, experimental studies are needed to
   elucidate their mechanisms and biological functions in asthma. Second,
   the original data were derived from public databases, with training and
   validation datasets from different sources. Further large-scale studies
   are required to explore the clinical relevance of these genes in
   asthma. Lastly, while KU-55933 was predicted as a potential therapeutic
   molecule, its lipophilic properties may limit clinical application, and
   its role in asthma requires validation through animal studies.
   Therefore, in subsequent studies, we will further investigate the
   specific diagnostic efficacy of AGPS and APELA in asthma animal models,
   as well as the therapeutic intervention effects of KU-55933, aiming to
   provide additional insights for effective asthma treatment.

Conclusion

   In summary, we identified two ferroptosis-related characteristic genes,
   AGPS and APELA, and developed a diagnostic model for asthma.
   Additionally, we analyzed the correlation between these characteristic
   genes and immune cell infiltration, shedding light on their role in
   asthma pathogenesis. We further identified KU-55933 as a potential
   therapeutic small molecule and explored its possible mechanisms in
   treating asthma by regulating ferroptosis. These findings provide new
   insights into the diagnosis and treatment of asthma, highlighting
   ferroptosis as a promising avenue for future research and clinical
   application.

Supplementary Information

   [190]Supplementary Information 1.^ (15KB, pdf)
   [191]Supplementary Information 2.^ (700.4KB, png)
   [192]Supplementary Information 3.^ (1MB, png)
   [193]Supplementary Information 4.^ (3.5KB, txt)
   [194]Supplementary Information 5.^ (23.6KB, zip)

Acknowledgements