A Machine Learning-Based Classification of Immunogenic Cell Death Regulators and Characterisation of Immune Microenvironment in Acute Ischemic Stroke

Immunogenic cell death (ICD) regulators exert a crucial part in quite a few in numerous biological processes. This study aimed to determine the function and diagnostic value of ICD regulators in acute ischemic stroke (AIS). 31 significant ICD regulators were identified from the gene expression omnibus (GEO) database in this work (the combination of the GSE16561 dataset and the GSE37587 dataset in the comparison of non-AIS and AIS patients). The random forest model was applied and 15 potential ICD regulators were screened to forecast the probability of AIS. A nomogram, on the basis of 11 latent ICD regulators, was performed. The resolution curve analysis indicated that patients can gain benefits from the nomogram. The consensus clustering approach was applied, and AIS patients were divided into 2 ICD clusters (cluster A and cluster B) based on the identified key ICD regulatory factors. To quantify the ICD pattern, 181 ICD-related dissimilarly expressed genes (DEGs) were selected for further investigation. The expression levels of NFKB1, NFKB2, and PARP1 were greater in gene cluster A than in gene cluster B. In conclusion, ICD regulators exerted a crucial part in the progress of AIS. The investigation made by us on ICD patterns perhaps informs prospective immunotherapeutic methods for AIS.


Introduction
Data from the Global Burden of Disease (GBD) in 2019 indicate that, after ischemic heart disease, stroke is the second leading cause of tertiary death (11.6% of total deaths) [1].AIS is the most common form of stroke, and its incidence has been on the rise in recent years [2].Te infammatory cascade triggered by cell death plays a signifcant role in the pathogenesis of AIS [3].Te levels of cytokines and oxidative stress markers also drive infammation, such as IL-1 and IL6 [4][5][6].In AIS, hypoxia activates HIF-1α, enabling HIF-lα the primary pathway to regulate angiogenesis after ischemia [7].Matrix metalloproteinase-9 (MMP9) and interleukin (IL6, IL8, and IL10) are frequently associated to the prognosis of AIS [8].Circulating immune cells and brain immune cells play a crucial dual function in the breakdown of the blood-brain barrier following AIS [9].
Immunogenic cell death (ICD) was recognized as a form of regulating cell death (RCD) [10].Tis procedure induces an adaptive immune response in the face of necrosis or predetermined death.Numerous in-depth investigations in the clinical utility of ICD have been performed recently.ICD stimulation results in the production of primitive antigenic epitopes and the release of damage-associated molecular patterns (DAMPs) by dying cells [11].DAMPs can bind to antigen-presenting cells, for instance, dendritic cells (DC), detect and phagocytose dead cell antigens, and deliver them to T cells to excite the adaptive immune response [12].Synergistic augmentation of immunogenic cell death and macrophage transformation is a novel cancer treatment combo [13].A novel genetic profle of aneurysms based on ICD features in non-neoplastic illness suggests that ICD patterns and the immune microenvironment are strongly associated with aneurysms [14].Earlier investigations have merely discovered the value of a modest number of immune cells and immunological-related chemicals in AIS.Nevertheless, AIS's ICD perspective is wanted.
Terefore, we systematically investigated ICD regulators in AIS in this work.Immunoassays between normal tissue and AIS blood samples, as well as a thorough analysis of several subtypes of AIS, will reveal the alterations in the occurrence of ICDs and their associated genes.We developed a gene model to forecast AIS susceptibility on the basis of 11 candidate ICD regulators and observed that patients were capable to get substantial advantages in the model.We made a comparison of biological functions in light of the fact that the two clusters share distinct immunological properties.We found that ICD modifcation mode exercised a substantial efect on AIS.It would provide brand new information to the investigation into the pathogenic mechanism of AIS.

Materials and Methods
2.1.Data Gathering.Te GSE16561 combined GSE37587 dataset of 24 healthy adult subjects and 107 AIS patients were obtained from the GEO database (https://www.ncbi.nlm.nih.gov/geo).Clinical data for both datasets can be found in previous studies [15,16].All patients from the two datasets met the following criteria: age ≥18 years, MRI diagnosis of AIS, and blood drawn within 48 hours of onset of stroke symptoms.Te datasets were chosen to elucidate gene expression in peripheral whole blood from patients with acute ischemic stroke in order to identify a set of genes for the diagnosis of acute ischemic stroke.Moreover, the data structure and characteristics of both datasets are identical.All patients met the following criteria: age ≥18 years, MRI diagnosis of AIS, and blood drawn within 48 hours of onset of stroke symptoms.Consequently, we combined and normalized the gene expression matrices of the two datasets to investigate the role of ICD-associated genes in acute ischemic stroke.31 ICD moderators in the terminal normal dataset were annotated [17]: ATG5, BAX, CALR, CASP1, CASP8, CD4, CD8A, CD8B, CXCR3, ENTPD1, FOXP3, HMGB1, HSP90AA1, IFNA1, IFNB1, IFNG, IFNGR1, IL10, IL1B, IL1R1, IL6, LY96, MYD88, NLRP3, NT5E, P2RX7, PDIA3, PIK3CA, PRF1, TLR4, and TNF.

Variations in ICD Regulators among Various Samples and
Associated Analysis.We analyzed the variation in the gene expression between normal and AIS samples by applying the "limma" package.We used Spearman's rank association analysis to determine the relationship of the expression of ICD regulators in AIS.

Te Establishment of a Random Forest Model and a Nomogram Model.
Random forest (RF) and support vector machine (SVM) models were developed by applying the random forest software package as training models to forecast the happening of AIS."PROC" package performed receiver operating characteristic (ROC) curve as well.We used a ten-fold cross-validation curve to make an estimation of the predictive quality of the RF model for ischemic stroke.Te red line stood for the experimental group's error, the black line was on behalf of the error of all samples, and the green line represented the error of the control group.Tereafter, we analyzed the importance of 15 ICD regulators and selected 11 suitable ICD regulators.On the basis of the 11 ICD regulators that were selected, the nomogram model was established by applying the "rms" package.Te calibration curve was employed to evaluate the congruence between the anticipated and real values separately.Decision curve analysis (DCA) was carried out, and a clinical efect curve was produced to determine whether the decisions on the basis of the model were benefcial to the patient.

Identifcation of Molecular Subtypes on the Basis of the
Momentous ICD Regulators.Consensus clustering is a kind of technique, which is applied to identify every member and its subgroup fgure, as well as to validate clustering rationale on the basis of resampling.Using the "Consensu-sClusterPlus" package in R [18], the consensus clustering method was employed to fnd ICD-related patterns on the basis of the signifcant ICD regulators.

Identifcation and Functional Enrichment of Variously
Expressed Genes in Various ICD Models.Te "limma" package was employed in R and variously expressed genes among diverse ICD patterns were recognized (p < 0.01).We employed principal component analysis (PCA) algorithms to calculate out the ICD score for every sample to quantify the ICD patterns.Trough applying the "clusterProfler" package, gene ontology (GO) functional annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis were made to fnd out the latent mechanism of variously expressed genes.Gene set enrichment analysis (GSEA) was also performed by the "clusterProfler" package.Te MSigDB C2 set was applied as the reference gene set.

Evaluation of Immune Cell
Infltration.We used the "ESTIMATE" package [19] to measure the immune cell infltration score in every sample.We made heat maps on account of the gene expression and immune scores of the samples and analyzed the distinctions in immune scores in cluster A and cluster B as well.In addition, we also examined the correlations between signifcant ICD regulators and immunological ratings.

Expression Landscape of ICD Regulators among Various
Samples.Te participation covered in this survey ranged from thirty-one various ICD regulators.Figure 1(a) highlighted the AIS samples and the normal samples' expression levels of 15 genes' substantial variations.Tese genes covered CASP1, CASP8, CD8A, CD8B, CXCR3, ENTPD1, IFNGR1, IL1R1, LY96, MYD88, NLRP3, P2RX7, PIK3CA, TLR4, and TNF.We employed a heat map to depict the distribution of 15 key ICD regulators among the samples that were estimated (Figure 1(b)).It demonstrated each location of 15 regulators in the chromosomes, which was conducted by the "RCircos" package (Figure 1(c)).

Correlation between Diferent ICD Regulators in AIS.
In order to study the associate degree that exists among ICD regulators working in various AIS, we analyzed correlation coefcients between ICD regulators based on gene expression and plotted a heat map (Figure 2(a)).Our investigation revealed that IL10, CXCR3, HSP90AA1, CD80A, NT5E, HMGB1, PIK3CA, ATG5, and PRF were related with each other strongly.A strong positive association can also be observed among NLRP3, IFNGR1, MYD88, TLR4, IL1B, and ENTPD1.In addition, substantial positive associations were discovered among four distinct pairs of genes (Figure 2(b)).

Establishment of the RF Model and SVM Model.
In order to provide a precise forecast regarding the occurrence of AIS, an RF was established and SVM was selected to choose candidate ICD regulators from the 15 ICD regulators.RF model is a random forest model that generates a fnal result based on the output of multiple decision trees generated at random.Te RF model has a potent capacity for capturing global data characteristics, a strong capacity for model generalization, and the capacity to parallelize calculations swiftly.SVM model is a common classifcation model that is suited for small samples with clear classifcation boundaries to determine the optimal segmentation plane.Terefore, these two models are an excellent ft for our investigation.Boxplots of residual, as shown in Figure 3(a), manifested the fact that the RF model's residuals were minimal.Te residuals of majority of model cases were correspondingly tiny.Terefore, the RF model was selected as the optimal model to forecast the occurrence of AIS.We established the ROC curve to appraise the model, and its AUC value suggested that the RF model enjoyed more precision than the SVM  A sum of 181 ICD-related DEGs were chosen for prospective investigation between the two ICD patterns.GO functional annotation and KEGG pathway analysis were made for a better comprehension of the latent mechanism behind these DEGs in AIS (Figure 5(h)).We discovered that the majority of the gene sets were enriched in the processes of T cell activation, leukocyte activation, regulation of immune efector, and diferentiation of lymphocytes and other processes.We discovered that the diferentially expressed genes became enriched in certain biological pathways by using GSEA analysis, for instance, the cytotoxic pathway, the lymphocyte pathway, the cell adhesion pathway, the T cell receptor pathway, the MHC and IL17 pathway, and so on (Figure 5(i)).It suggested that the immunological activity of cluster A was considerably greater than that of cluster B.
After that, we utilized "ESTIMATE" to compute the quantity of immune cells presented in the AIS samples, and we investigated the degree to which the 15 most signifcant ICD regulators which were correlated with the immune cells (Figure 6(a)).We compared the two ICD patterns and observed the distinctions in immune cell infltration.Neutrophils and eosinophils were more prevalent in cluster B than they were in cluster A, while T cells and MDSC were the opposite (Figure 6(b)).In addition, the link between four major ICD regulators and immune cell infltration was International Journal of Clinical Practice demonstrated (Figures 6(c)-6(f )).Tese fndings reafrmed that ICD alteration played a crucial regulatory function in the formation of distinct blood immunological microenvironments in AIS patients.

Discussion
Te transformation of nonimmunogenic cells into immunogenic cells to bring an immune reaction in the course of cell death is referred to as immunogenic cell death (ICD) [20].ICDs can be triggered by a variety of stimuli, such as viral infections, anthracyclines, certain types of radiation therapy, and photodynamic therapy [21].DAMPs, generated when cells are stimulated, can attach to pattern recognition receptors (PRRs) on the surface of DC cells, triggering a cascade of physiological events that ultimately activate innate and adaptive immune responses [22].Te ICD pattern was primarily discovered and examined in the feld of tumour therapy.Nevertheless, the role that ICD played in AIS was still not fully appreciated.
It is worth noting that infammation and immunological pathways play a signifcant role in the pathophysiology of the onset, acute damage cascades, and chronic course of ischemic stroke [23][24][25].Te mechanism of secondary injury after ischemia may be due to the generation of intracerebral infammation after ischemic stroke, which accelerates the formation of ischemic injury and afects neuronal mortality and nerve tissue regeneration [26].Neuroinfammatory response after cerebral ischemia is characterized by activation of microglia, activation of astrocytes, and increase of infammatory bodies.Malignant edema and hemorrhagic transformation are the most common clinical symptoms, and their mechanisms have been characterized in detail in animal models [27].A plasma exosome (CircOGDH) has been recognized as a therapeutic target and penumbral biomarker for acute ischemic stroke [28].Consequently, diagnostic and prognosis evaluation on the basis of peripheral blood biomarkers will exert a signifcant efect on the mortality control of AIS patients.We believe that our research will contribute to a greater understanding of the crucial role of ICD-related genes and pathways and provide novel diagnosis, prevention, and immunotherapy for stroke patients as well.6 International Journal of Clinical Practice could be advantageous to the AIS.CD8B, P2RX7, IFNGR1, TLR4, ENTPD1, and CD8A were protective factors, while NLRP3, MYD88, IL1R1, PIK3CA, and LY96 were hazard elements for AIS.
As an essential component of innate immunity, NLRP3 infammasome plays a crucial role in the immunological response of the body and the development of illness [29].It can be triggered by diverse infections or danger signals.A prior work employing bioinformatics and in vivo experiments confrmed that the suppression of IL1R1 or CASP4 ameliorated pyroptosis triggered by NLRP3 infammasomes [30].MYD88 performs a critical signal transduction role in innate and adaptive immune responses.Recent research has demonstrated that mesencephalic astrocyte derived neurotrophic factor (MANF) inhibits the production of proinfammatory factors and relies on the TLR4/MyD88/NF-B pathway to maintain the integrity of the blood-brain barrier in a geriatric mouse model following an ischemic stroke [31].
Te connection between ICD regulators and AIS immunological features was studied subsequently.Te expression of various immune response gene sets and infltrating immune cells was investigated by the means of using GSEA, ESTIMATE, GO, and KEGG analyses.Tese immunological characteristics were found to be tightly associated with ICD regulators, which represented that ICD was essential in controlling the blood immune milieu of AIS.Two clusters with distinctive ICD patterns were discovered on the basis of the expression profle of core ICD regulators and ICD-associated DEGs in AIS.Each cluster possesses its unique immunological properties.For example, cluster A patients had a higher proportion of T cells in their blood.Te classifcation of immunological clusters contributes to the  Te association between ICD regulatory patterns and signifcant immune biomarkers of AIS was also examined.Te pathological process of AIS is overwhelmingly tanglesome, including cell excitotoxicity, oxidative stress, cell death processes, and neuroinfammation [32].Simultaneously, a great number of neurotoxic or neuroprotective signalling pathways are intricately involved in the aforementioned pathophysiological processes.In addition, these signalling pathways have therapeutic potential, as targeting them was likely to be a therapeutic approach for ischemic strokes.Te expression features of cytokines regulating infammatory responses and proteins involved in angiogenesis in the brain and peripheral circulation gave a highlight of the need for the identifcation of original biomarkers in the circulatory system [33].Nuclear factor-κB (NF-κB) signalling pathway was essential for maintaining the blood-brain barrier's integrity and therefore was used as a therapeutic target for AIS [34].As a member of the sirtuin family, SIRT1 regulated a broad physiological process, covering apoptosis and infammatory reaction, and may be protective factors for stroke [35].Poly (ADP-ribose) polymerase-1 (PARP-1) regulated cell apoptosis and tissue necrosis in AIS and was associated with prognosis [36].Te fndings demonstrated  International Journal of Clinical Practice a fact that NFKB1, NFKB2, and PARP1's expression levels were higher, but the expression levels of SIRT1 were lower in gene cluster A than in gene cluster B, which suggested that gene cluster A was strongly correlated with AIS.
Tis study assessed the function of ICD regulators in patients with AIS.Tis study demonstrated that ICD regulators can easily diferentiate AIS patients from wholesome controls.Two distinct ICD clusters were identifed according to 15 ICD regulators, and the model was enhanced by ICDrelated DEG expressions, which helped to discovering feasible predictive indicators for the therapy of AIS.ICD expression, immune scores, and biological functional pathways were signifcantly diverse between the two ICD clusters of AIS.Tese fndings can bring innovative immunotherapeutic concepts for AIS.
Nevertheless, there are certain limitations to the study.First, the data were obtained from a limited sample size GEO data set.It may take a considerable amount of time to collect a great number of samples to get a total comprehension of ICD in AIS.Second, we did not conduct experimental verifcation because of the difculties in the process of acquiring AIS samples.

Conclusion
In conclusion, the study identifed 15 potential ICD regulators and a nomogram model, which was capable to forecast the prevalence of AIS with accuracy.Tis could have signifcant implications for clinical screening of AIS susceptibility genes and disease course monitoring.Moreover, we discovered signifcant disparities between the two ICD modes in the blood immune microenvironment.Tese fndings can guide the development of diagnosis and individualized immunotherapies for AIS patients.

model (Figure 3 3 . 4 .
(b)).Te ten-fold cross-validation curve uncovered a truth that the RF model was most accurate.Nevertheless, we depicted 11 top signifcant genes of the 15 ICD regulators after these genes were ranked by the means of their signifcance (Figure3(d)).Establishment of the Nomogram Model.On the basis of 11 candidates for the ICD regulations, a nomogram model was created to estimate the prevalence of AIS (Figure4(a)).CD8B, P2RX7, IFNGR1, TLR4, ENTPD1, and CD8A were protective factors, while NLRP3, MYD88, IL1R1, PIK3CA, and LY96 were hazard elements for AIS.Te nomogram model's predictivity was appeared to be accurate by applying calibration curves (Figure4(b)).Te DCA curve demonstrated that judgments, on the basis of the nomogram model, may be benefcial to AIS patients (Figure 4(c)).Te clinical infuence curve uncovered the nomogram model's outstanding prediction potential capacity (Figure 4(d)).

3. 5 .
Signifcant ICD Regulators Identifed Two Distinct ICD Patterns.Te "ConsensusClusterPlus" program was utilized to identify two unique ICD patterns on the basis of 15 major ICD regulators by employing the consensus clustering technique (Figures 5(a)-5(d)).15 key ICD regulators' expression levels were compared in the two clusters, and afterward a heat map and histogram were generated to indict the variations (Figure 5(e)).Tere were discernible variations between cluster B and cluster A in the expression level of CD8A, CD8B, CXCR3, ENTPD1, IFNGR1, LY96, and TLR4 (Figures 5(e) and 5(f )).Te PCA results revealed that the 15 key ICD regulators were able to difer between the two ICD patterns totally (Figure 5(g)).
Patterns in Distinguishing AIS.Te consensus clustering method was utilized to classify AIS patients into distinct genomic subgroups on the basis of the 181 ICD-associated DEGs for the prospective validation of the ICD patterns (p < 0.05) (Figures 7(a)-7(d)).Two unique ICD gene patterns were identifed, which covered gene cluster A and gene cluster B.

Figure 2 :Figure 3 :
Figure 2: Correlations among 31 ICD regulators.(a) Heat map of association for 31 ICD regulators.(b-e) Four pairs of ICD supervisors with the highest correlation and scatter plots of their correlation (Spearman rank coefcient r > 0.7).

Figure 4 :Figure 5
Figure 4: Establishment of the nomogram model.(a) Te establishment of the nomogram model on account of the 11 alternative ICD regulators.(b) Calibration curve reveals the predictive capability of the nomogram model.(c) AIS patients may gain advantages from decisions on account of the nomogram.(d) Clinical infuence of the nomogram model as estimated by using the clinical impact curve.

Figure 5 :Figure 6 :
Figure 5: Consensus clustering of the 15 vital ICD regulators in AIS.(a-d) Consensus matrices of the 15 signifcant ICD regulators for k � 2-5.(e) Expression heat map of the 15 crucial ICD regulators in cluster A and cluster B. (f ) 15 vital ICD's distinctive expression of regulators in cluster A and cluster B ( * p < 0.05, * * p < 0.01, and * * * p < 0.001).(g) Principal component analysis recovers transcriptomes' striking distinctions in the two ICD patterns based on the expression profles of 15 essential ICD regulators.(h) GO and KEGG analyses investigate the possible mechanism based on the infuence of the 181 ICD-associated DEGs on the happening and progression of AIS.(i) GSEA analysis examines the latent mechanism based on the efect of ICD-related DEGs on the development and progression of AIS.

Figure 7 :
Figure 7: 181 ICD-associated DEGs' consensus clustering in AIS ( * p < 0.05, * * p < 0.01, and * * * p < 0.001).(a-d) 181 ICD-associated DEGs' consensus matrices for k � 2-5.(e) Expression heat map of the 181 ICD-related DEGs in gene cluster A and gene cluster B. (f ) 15 essential ICD regulators' distinctive expression histogram in gene cluster A and gene cluster B. (g) Diferential immune cell infltration between gene cluster A and gene cluster B. (h) Te relationship among ICD patterns, ICD gene patterns, and ICD scores was demonstrated by Sankey diagram.(i) AIS-associated markers' distinctive expression between gene cluster A and gene cluster B.