Abstract Despite advancements in trauma care, uncontrolled hemorrhage and trauma-induced coagulopathy (TIC) remain the leading causes of preventable deaths after trauma. Understanding the genetic underpinnings and molecular mechanisms of TIC is crucial for developing effective diagnostic and therapeutic strategies. This study employed a comprehensive bioinformatics approach to elucidate the genetic landscape associated with TIC. Gene expression data from 20 samples, comprising 10 controls and 10 severe trauma patients with TIC, were analyzed. This approach included principal component analysis, differential gene expression analysis using DESeq2, Gene Set Enrichment Analysis (GSEA), weighted gene co-expression network analysis (WGCNA), and machine learning (ML) algorithms (support vector machine-recursive feature elimination, least absolute shrinkage and selection operator, and random forest) for feature gene identification. Functional analysis of genes and immunoinfiltration analysis were also conducted. A total of 1014 differentially expressed genes (DEGs) were identified, indicating significant genetic alterations in TIC. GSEA confirmed the involvement of critical pathways, and WGCNA identified 35 relevant gene modules. The integration of ML algorithms highlighted nine key feature genes (TFPI, MMP9, ABCG5, TPSAB1, TK1, IGKV3D.11, SAMSN1, TIMP3, and GZMB). Immunoinfiltration analysis revealed distinct immune cell compositions in TIC samples. The multifactor regulation network provided insights into complex gene regulatory mechanisms. This study presents a detailed genetic and molecular profile of TIC. Integrating various bioinformatics tools and ML algorithms has enabled the identification of potential biomarkers and therapeutic targets. These findings could significantly contribute to improving the diagnostic accuracy and treatment efficacy for patients with TIC, potentially reducing the mortality rates associated with trauma. Supplementary Information The online version contains supplementary material available at 10.1038/s41598-025-10323-4. Keywords: Trauma-induced coagulopathy, Differential gene expression, Weighted gene coexpression network analysis, Machine learning algorithms, Biomarker identification, Immunoinfiltration analysis Subject terms: Clinical genetics, Sequencing Introduction Despite advancements in trauma care, uncontrolled hemorrhage coupled with coagulopathy continues to be a significant clinical challenge^[60]1 and is the leading preventable cause of death after polytrauma^[61]2. The onset of fatal bleeding is alarmingly swift, often occurring within a median time of 1.65 h after hospital admission^[62]3. Furthermore, upon arrival at the emergency department, 1 in 4 patients with severe trauma exhibit laboratory signs indicative of coagulopathy^[63]4. The exploration of trauma-induced coagulopathy (TIC) pathophysiology aims to improve management strategies for severely injured patients, particularly in the context of massive blood loss. Advancements in point-of-care diagnostics and a deeper understanding of the adverse effects of conventional treatments, such as intravenous fluids and fresh-frozen plasma (FFP), have spurred the development of novel approaches^[64]5. TIC primarily results from blood loss and the activation of the protein C system due to hypovolemia, leading to increased fibrinolytic activity. This condition is intensified by factors such as hemodilution, clotting factor and platelet consumption, hypothermia, acidosis, anemia, and hypocalcemia, all of which impair hemostatic function^[65]6. Randomized controlled trials focusing on trauma patients with hemorrhagic shock in the first 24 h after injury have revealed the impact of various interventions on bleeding, transfusion needs, TIC correction, and mortality^[66]7. There is an ongoing debate on whether TIC represents entities such as disseminated intravascular coagulation (DIC) with a fibrinolytic phenotype or acute coagulopathy of trauma shock, with studies such as Johansson et al. attempting to define these conditions through biomarker profiles^[67]8. Viscoelastic tests, such as rotational thromboelastometry or thrombelastography, provide comprehensive insights into trauma coagulation processes, supporting strategies for early, personalized, and goal-oriented treatment, such as the AUVA Trauma Hospital algorithm^[68]9. A clear understanding of TIC, thrombosis, and hemostasis after trauma is essential, as is recognizing DIC pathogenesis in the early stages of trauma^[69]10. Despite the upregulation of procoagulant mechanisms, a significant portion of trauma patients exhibit laboratory signs of TIC, correlating with increased mortality and worse outcomes^[70]11. Literature reviews have recently delved into advancements in TIC comprehension, highlighting the roles of platelet dysfunction, endothelial activation, and fibrinolysis^[71]12. However, the clinical translation of these laboratory findings into bleeding complications remains ambiguous, presenting challenges in diagnosing and treating TIC^[72]13. The complexity of TIC, marked by initial hypocoagulability and subsequent hypercoagulability due to trauma-induced physiological responses, continues to be a leading preventable cause of death among trauma patients. The investigation focused on discerning individual effects of acidosis and hypothermia on coagulopathy using established clinical and bedside tests, such as bleeding and prothrombin time (PT), with clinical coagulopathy indicated by prolonged splenic bleeding time due to hypothermia achieved by a cold water-circulating blanket^[73]14. The proteomic analysis of Crotalus durissus terrificus venom expanded the known toxin profile, finding unique components not previously reported in the venom gland transcriptome of C. durissus collilineatus^[74]15. A multivariate analysis examining risk factors for acute traumatic coagulopathy highlighted shock at the scene and in the emergency department as significant, tripling the risk of developing coagulopathy^[75]16. The RiboMinus™ technique is a breakthrough for isolating RNA devoid of large rRNA, optimizing the purity of whole transcriptome sequencing^[76]17. Whole transcriptome analysis was explored for its critical role in interpreting genomic functions, mapping genetic networks, and identifying biomarkers responsive to disease, with a discussion encompassing shotgun and targeted sequencing approaches^[77]18. The correlation between sympathetic nervous system markers, such as catecholamines, and biomarkers indicative of coagulopathy, endotheliopathy, and inflammation was established through partial least squares-discriminant and regression analyses, with a marked covariant relationship observed at admission and 24 h after injury^[78]19. mRNA sequencing research uncovered that acrolein potentially triggers autophagy via the Akt/mammalian target of rapamycin pathway, implicating it in traumatic brain injury-related coagulopathy and vascular leakage and pointing to a novel therapeutic avenue^[79]20. In hematology, transcriptome sequencing proves its diagnostic value in acute promyelocytic leukemia cases by identifying the fip1l1::rara fusion gene, particularly when conventional diagnostic techniques fall short. The same study also differentiated gene expression profiles in comparative analyses with other myeloid malignancies and acute lymphoblastic leukemia^[80]21. A transcriptome meta-analysis of coronavirus disease 2019 revealed genes associated with coagulopathy, lung fibrosis, and multiorgan damage, providing insights into the long-term sequelae of the disease as evidenced in lung autopsy samples^[81]22. The TIC investigation discovered 1014 differentially expressed genes (DEGs), providing insights into molecular alterations after severe trauma. This analysis discriminated between 711 upregulated and 303 downregulated genes, highlighting intricate regulatory mechanisms. Through Gene Set Enrichment Analysis (GSEA), pathways such as interleukin (IL)−17 signaling and complement and coagulation cascades were identified as significantly enriched, underscoring their potential as therapeutic targets in TIC. Weighted gene coexpression network analysis (WGCNA) allowed for the delineating 35 gene modules, offering fresh perspectives on gene interactions and suggesting new biomarkers or therapeutic avenues. The gene identification process was further refined by incorporating machine learning (ML) algorithms—support vector machine-recursive feature elimination (SVM-RFE), least absolute shrinkage and selection operator (LASSO), and random forest (RF)—elevating the precision of this research. Functional analyses using Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways clarified the roles of these genes in key biological processes, paving the way for targeted interventions in TIC treatment. This study not only supports but also extends the current knowledge base, leveraging a blend of WGCNA, ML, and functional analytics to enhance the understanding of TIC. This methodological synthesis places our research at the cutting edge of TIC studies, providing a comprehensive view that could inform future preventive and therapeutic strategies. Methods Study design The study selected 20 patients diagnosed with severe multiple injuries (injury severity score > 16) who were hospitalized in the Trauma Department of Chongqing University Central Hospital from January to September 2023. Patients were divided into two groups of 10 each based on whether they developed TIC or not (TIC patients and control groups). Inclusion criteria were as follows: met the diagnostic criteria for acute TIC and had complete information and authorized representatives, activated partial thromboplastin time (APTT) > 60 s, PT > 18 s, and international normalized ratio (INR) > 1.5^[82]23. Exclusion criteria included patients who died before reaching the hospital or in the emergency room, those with acute or chronic liver dysfunction rated as Child-Pugh B or C, those treated with anticoagulants or antiplatelet drugs before the injury, transfers from other hospitals or arrivals at the emergency room > 90 min after the injury, and cases with missing data deemed essential by the research team. Patients and families who were uncooperative with the treatment were also excluded. The study adhered to the ethical guidelines outlined in the Declaration of Helsinki. Approval for the study protocol was granted by the Ethics Committee of the Chongqing University Central Hospital (Approval No.: Scientific Research 2022 [74]). Informed consent was obtained in writing from all participants who enrolled in the study. Participants agreed to the sharing of their deidentified information. Flow chart of this study was demonstrated in Fig. [83]1. Fig. 1. [84]Fig. 1 [85]Open in a new tab Flow diagram of the methodologies used to investigate the biological characteristics of TIC. TIC, trauma-induced coagulopathy. Clinical variable analysis Using GraphPad Prism (GraphPad Software, Inc., San Diego, CA, USA), descriptive statistics were computed for all demographic and clinical variables. To assess the differences in the demographic and clinical characteristics between the groups, χ^2 test and independent samples t-test were conducted. Sample processing and RNA sequencing Blood samples for the study were collected in PAXgene blood RNA tubes (PreAnalytix/Qiagen) through an existing intravenous catheter. These samples were stored at 80 °C until ready for batch processing. Approximately 30 mg of the samples ground in liquid nitrogen were processed using Trizol reagent (Invitrogen, USA) to extract the total RNA. RNA purity was assessed using NanoDrop, whereas RNA concentration was measured using the Qubit RNA Broad-Range Assay Kit. RNA integrity was evaluated using RNA Screen Tape. Based on the total amount of RNA, a quantity ranging from 0.1 to 1 µg was isolated using the NEBNext^® Poly(A) mRNA Magnetic Isolation Module. Isolated mRNA was used to construct mRNA libraries with the NEBNext^® Ultra™ II mRNA Library Prep Kit for Illumina^®. The library concentration was determined using the Qubit™ dsDNA HS Assay Kit, and library fragment distribution was assessed with D1000 Screen Tape. The precise molar concentration of the library was quantified using the KAPA Library Quant Kit (Illumina) universal quantitative polymerase chain reaction (qPCR) mix. Finally, library sequencing was conducted on the NovaSeq 6000 System using the NovaSeq S4 Reagent Kit. The sequencing experiment was performed at the HaploX Genomic Center. Raw data were deposited in the Gene Expression Omnibus database ([86]https://www.ncbi.nlm.nih.gov/geo/). Identification of DEGs and principal component analysis (PCA) PCA was conducted to evaluate the variance between the TIC and control groups. For this analysis, two principal components were extracted. DEGs between the control and TIC samples were analyzed using the “DESeq2” package in R^[87]24. DEG threshold was set at an absolute log[2]Fold change |log2FC| > 1 and p < 0.05. DEGs were visually represented using a volcano plot. The “ggplot2” and “pheatmap” packages in R (version 4.3.2) were employed to create the clustering heatmap of all DEGs, providing a comprehensive view of gene expression changes between the groups. GSEA To provide a more intuitive understanding of gene expression levels in significantly enriched functional pathways, GSEA was conducted using R software. Furthermore, to elucidate the relationship between signature genes and signaling pathways in TIC, the cohort was stratified based on the median expression level of hub genes. GSEA was conducted across these different subgroups, focusing on identifying significant associations with p[adj] < 0.05^[88]25. Identification of modules by constructing the gene coexpression network Using “WGCNA” package in R, a gene coexpression network for a severe trauma patient with TIC was constructed. Initially, 32,796 genes were selected from the training group. These genes were subjected to a similarity matrix computation converted into an adjacency matrix using a soft threshold power (β) between 1 and 20, determined by the “pickSoftThreshold” function. This step established a scale-free topology graphically validated by the relationship between log10(k) and log10[p(k)]. After this, the adjacency matrix was transformed into a Topological Overlap Matrix, and the modules within this matrix were identified using functions, such as “pickSoftThreshold” and “plotDendroAndColors.” These modules were analyzed for their correlation with all clinical characteristic traits, identifying the ME blue module as particularly significant^[89]26. Functional and pathway enrichment analysis and visualization of intersecting genes A Venn diagram analysis was conducted to identify intersecting genes among DEGs, TIC-related genes, and modules derived from WGCNA. To elucidate the functions of intersecting genes, GO annotation^[90]27 and KEGG enrichment analysis^[91]28 were performed using the “clusterProfiler” package. These analyses were conducted to gain insights into the biological roles of intersecting genes. Each gene’s fundamental function was inferred based on its protein domain characteristics and the existing research literature. This approach provided a preliminary understanding of the gene’s functional relevance. GO and KEGG are comprehensive databases that categorize gene functions based on different classification criteria. GO annotation categorizes gene functions into three main levels: Biological Process (BP), Cellular Component (CC), and Molecular Function (MF). This classification system helps discern the primary roles of DEGs within these categories. In addition to annotating human gene pathways, the human gene pathway database was also consulted. In contrast, KEGG is a database that focuses on pathways related to genes. Adjusted p-values (p[adj]) < 0.05 were considered statistically significant, indicating that the identified functional annotations and pathway enrichments were highly likely to be biologically meaningful which was generated by DAVID^[92]29. To construct a protein interaction network map involving the intersecting genes, the 81 intersecting genes were inputted into the STRING database^[93]30, leading to the generation of a protein-protein interaction (PPI) network ([94]https://cn.stringdb.org). This PPI network (version 3.9.0) was visualized and constructed using Cytoscape^[95]31. Verification and screening of diagnostic markers This study employed a combination of ML techniques to identify key genes for diagnosing TIC. Initially, a set of genes was selected and refined using several advanced methods. First, SVM-RFE was utilized. This technique is a supervised ML for either regression or classification, necessitating a training set with labels^[96]32. SVM-RFE specifically helps narrow the feature set by training a subset of features from different categories, highlighting the most predictive features. Next, the “glmnet” package in R was used to conduct LASSO regression. LASSO regression is particularly effective for computing linear models and retaining essential variables. It uses minimum absolute shrinkage, which helps in variable selection and regularization to enhance prediction accuracy. The variables following a binomial distribution were incorporated into the LASSO classification. This approach was augmented with a 1 standard error (SE) λ-value based on the 1–SE criterion, which is known to build models with robust performance while using only a limited number of cross-validation variables^[97]33. Finally, the RF algorithm played a crucial role. It was employed to rank the genes based on their importance. Genes with a relative importance value exceeding 0.25 were considered significant because this threshold typically indicates more than just a random chance occurrence^[98]34. The intersection of LASSO logistic regression, SVM-RFE, and RF results was used to pinpoint the most significant feature genes for the study. This integrative approach ensured a comprehensive and reliable selection of genes pertinent to TIC diagnosis. Diagnostic column line graph construction A graphical representation facilitating rapid and approximate complex calculations^[99]35 or a nomogram was utilized in this study. The nomogram model was constructed using the rms package, incorporating selected gene signatures from the diagnostic model. This model aimed to predict TIC in severe traumatic patients. To evaluate the nomogram’s effectiveness, the concordance index was calculated using a bootstrap method with 1000 resamples to measure the model’s discrimination ability. Calibration curves were plotted to compare the nomogram’s predicted probabilities with actual observed rates. Decision curve analysis (DCA) curves were also employed. DCA curves are beneficial in assessing a model’s clinical utility, providing an integrated view of the benefits and potential harms associated with the prediction model, thus addressing the shortcomings of traditional statistical metrics^[100]36. Curve analysis of the receiver operating characteristic (ROC) curve ROC curves were constructed to evaluate the diagnostic potential of candidate genes and the columnar maps for TIC, which enabled to calculate the area under the curve (AUC) along with a 95% confidence interval (95% CI), providing a quantitative assessment of their diagnostic value. AUC > 0.70 was regarded as indicative of the optimal diagnostic utility. To enhance the clarity and comprehension of findings, the results of these analyses were visualized using R (version 4.3.2) software. This approach allowed for a more detailed and nuanced understanding of the diagnostic relevance of genes and maps under study^[101]37. Immunoinfiltration analysis All samples were analyzed using various immunoinfiltration methods, including XCELL, QUANTISEQ, MCPCOUNTER, EPIC, and CIBERSORT. CIBERSORT, a computational method, was employed to analyze the immune cell composition in TIC cases and control groups. This method, which leverages tissue gene expression profiles, is adept at identifying varying proportions of immune cells. For a detailed immunoinfiltration analysis, the “CIBERSORT” package in R was used^[102]38. To effectively present these findings, bar graphs were created to illustrate the proportion of each immune cell type in different samples. A comparative analysis of the immune cell composition between the TIC and control groups was also conducted using violet for visual representation. Furthermore, heatmaps depicting the correlation among the 22 infiltrating immune cells were created to provide a comprehensive view of the immune landscape. These heatmaps were generated using the “corrplot” package in R, which offers a clear and detailed visualization of the intricate immune cell interactions in the study samples^[103]39. The linkET package in R was utilized to create correlation heatmaps, linking the selected feature genes with immunoanalysis results obtained from CIBERSORT. MiRNA and transcript factors of the diagnostic gene network construction This study constructed a multifactor regulation network to understand the intricate interactions involving hub genes. To predict the transcription factors (TFs) and microRNAs (miRNAs) associated with these hub genes, two databases were utilized: NetworkAnalyst and miRNet^[104]40,[105]41. Hub genes with their respective miRNAs and TFs were integrated into a comprehensive regulatory network. The interaction between hub genes and potential TFs was based on the JASPAR database. This network was visualized using Cytoscape, providing a clear and detailed representation of the complex regulatory relationships. This approach allowed exploring the multifaceted regulatory mechanisms in this study’s context. RNA isolation and real-time qPCR analysis of biomarker genes To validate the bioinformatics analysis results, whole-blood samples from two groups of patients were collected: TIC patients and control controls. Total RNA was extracted from whole blood using Trizol reagent. These RNA samples were reverse transcribed into cDNA using the PrimeScript RT reagent kit and amplified with the TB Green Premix Ex Taq II kit (TaKaRa, Japan). The primer sequences for mRNA amplification are provided in Table [106]1. All mRNA primers were procured from Thermo Fisher Scientific (USA). To ensure consistency, gene expression levels were normalized to glyceraldehyde 3-phosphate dehydrogenase and calculated using the 2^−ΔΔCt method. Statistical significance was defined as p < 0.05. Primers used in this work can be seen in Table [107]1. Table 1. Forward and reverse primers of TFPI, MMP9, ABCG5, TPSAB1, TK1, SAMSN1, and TIMP3. Gene name Forward primer Reverse primer TFPI ATGGAACCCAGCTCAATGCT GGCACGACACAATCCTCTGT MMP9 TCTATGGTCCTCGCCCTGAA CATCGTCCACCGGACTCAAA ABCG5 GTGTAGTGGCTCTGCTGTCC GTGTAGTGGCTCTGCTGTCC TPSAB1 GCCCATACTGGATGCACTTC GAACTGTGGGTGCACGATGA TK1 TCATAGGCATCGACGAGGGG CACCTCGACCTCCTTCTCTGT TIMP3 CTGCTACTACCTGCCTTGCT GCAGGCGTAGTGTTTGGACT SAMSN1 CAAAAACCAAAGCGAAGCAGC GGTCCCAATCACAGGGTCAC GAPDH GTCAAGGCTGAGAACGGGAA AAATGAGCCCCAGCCTTCTC [108]Open in a new tab Results Participant demographics and clinical variables Among participants in the clinical trial with complete records, 10 individuals met the criteria for controls and 10 individuals met the criteria for TIC patients. Table [109]2 summarizes the patients’ demographic and clinical characteristics. Several clinical variables showed statistically significant findings, including APTT (p < 0.01), PT (p < 0.01), INR (p < 0.01), fibrinogen (p < 0.05), hemoglobin (p < 0.05), hematocrit (p < 0.05), ionized calcium (p < 0.05), lactic acid (p < 0.05), albumin (p < 0.05), and shock index (p < 0.01). Table 2. Clinical variable analysis between controls and TIC patients. Variables Trauma induce coagulopathy Normal 95% confidence interval t value P value Gender, n male/female 8/2 8/2 Age (mean ± STD) 49.7 ± 14.3 50.5 ± 18.4 −15.55, 17.15 0.103 0.919 Activated partial thromboplastin time (s) 62.4 ± 24.6 34.7 ± 2.4 −44.98, −10.32 3.352 0.004** Thrombin time (s) 20.5 ± 6.0 16.63 ± 1.9 −8.286, 0.466 1.877 0.077 Prothrombin time (s) 23.3 ± 4.4 14.88 ± 1.3 −11.64, −5.198 5.491 < 0.001*** INR 2.1 ± 0.6 1.194 ± 0.1 −1.356, −0.554 5.008 < 0.001*** Fibrinogen (g/L) 1.2 ± 0.7 2.866 ± 1.3 0.617, 2.657 3.37 0.003** D-dimer (µg/L) 18.2 ± 11.1 10.623 ± 10.7 −18.34, 3.250 1.468 0.159 Hemoglobin (g/L) 78.1 ± 33.2 110.18 ± 24.8 2.932, 61.23 2.312 0.033* Platelats (10^9/L) 125.8 ± 98.2 151.4 ± 53.9 −52.86, 104.1 0.686 0.502 Hematocrit % 23.3 ± 9.7 33.85 ± 7.2 2.063, 19.04 2.611 0.018* Clotting time/R time (min) 9.4 ± 7.1 5.8 ± 0.7 −12.11, 4.866 0.939 0.368 Clot Formation Time/k-time (min) 4.2 ± 2.5 2.775 ± 0.5 −4.453, 1.670 1 0.339 α-angle (°) 48 ± 14.2 56.9 ± 4.3 −8.414, 26.21 1.131 0.282 Maximum amplitude (mm) 42.1 ± 16.1 50.375 ± 2.9 −11.16, 27.75 0.937 0.368 Elasticity at platelet-fibrinogen lysis (%) 10.1 ± 28.0 1.25 ± 1.0 −42.33, 24.60 0.583 0.572 Lysis 30 (%) 10.1 ± 28.0 1.05 ± 1.1 −42.53, 24.41 0.596 0.563 Clotting time 1 value −6.5 ± 6.8 −2.125 ± 0.8 −3.796, 12.61 1.183 0.262 Ionized calcium (mmol/L) 1.8 ± 0.3 2.011 ± 0.1 0.010, 0.436 2.2 0.041* Lactic acid (mmol/L) 9.03 ± 6.6 2.72 ± 1.0 −10.99, −1.626 2.83 0.011* pH 7.3 ± 0.1 7.366 ± 0.1 −0.107, 0.197 0.643 0.532 Albumin (g/L) 25.9 ± 9.0 35.09 ± 6.6 1.377, 17.06 2.47 0.024* Total bilirubin (µmol/L) 10.6 ± 5.4 12.74 ± 4.1 −2.611, 6.891 0.946 0.357 Alanine aminotransferase (u/L) 48.9 ± 47.5 39.8 ± 21.5 −45.59, 27.39 0.524 0.607 Aspartate aminotransferase (u/L) 116.3 ± 192.7 70.7 ± 43.6 −183.9, 92.74 0.692 0.498 Creatinine (µmol/L) 105.5 ± 37.5 89 ± 18.9 −45.91, 12.91 1.178 0.254 Temperature (°) 36.5 ± 0.3 30.568 ± 11.6 −8.981, 3.235 0.988 0.336 Shock index 1.6 ± 0.6 0.858 ± 0.2 −1.127, −0.257 3.343 0.004** Injury severity score 35.9 ± 11.8 32.2 ± 16.6 −17.99, 10.59 0.534 0.593 [110]Open in a new tab Identification of DEGs between the TIC and control groups In this study, gene expression matrices for the training group were generated following a series of preprocessing steps, including data cleaning, normalization, batch effect removal, and merging. This process involved data from two distinct groups: 10 samples from controls (forming the control group) and 10 samples from patients with TIC, specifically those who had suffered severe trauma. The PCA score plot (Fig. [111]2A) demonstrated a distinct and consistent clustering of the two groups based on the two principal components. This clear separation between the TIC and control groups validated the effectiveness of the preprocessing steps and indicated that these groups were suitably differentiated for further analysis, emphasizing the presence of significant between-group differences. The “DESeq2” package was employed for analysis and set specific criteria for identifying significant DEGs: p < 0.05 and a |log2FC| > 1 in absolute value. Under these criteria, 1014 DEGs were identified, with 711 upregulated and 303 downregulated genes (Fig. [112]2B). The image is a heatmap depicting the correlation between different genes. Gene names are listed on the left side and at the top of the heatmap. Correlation values are color-coded, ranging from 1 (blue for complete negative correlation) to + 1 (red for complete positive correlation). The intensity of the color indicates the strength of the correlation, with darker reds indicating strong positive and darker blues indicating strong negative correlations. The numbers at each intersection provide the exact correlation coefficient (Fig. [113]2C). A heatmap illustrates DEGs between the groups, visually representing the gene expression differences and highlighting the molecular alterations associated with TIC (Fig. [114]2D). Fig. 2. [115]Fig. 2 [116]Open in a new tab Data processing and identification of DEGs. (A) PCA score plot differentiating controls from severe injury patients with TIC. (B) Volcano plot showing DEGs. Orange dots, upregulated genes; dark-blue dots, downregulated genes. (C) Analysis of correlation coefficients, specifically focusing on the top 20 DEGs. (D) Heatmap of the top 50 upregulated and downregulated genes, illustrating differential expression. GSEA of all count data To verify the accuracy of the phenotypes of the samples collected, GSEA was conducted on raw count data. Results showed that the original counts of the samples were closely related to the complement and coagulation cascades, IL-17 signaling pathway, antigen processing and presentation, cell cycle, cytokine-cytokine receptor interaction, tumor necrosis factor (TNF) signaling pathway, phosphatidylinositol 3-kinase (PI3K)/Akt signaling pathway, and natural killer (NK) cell-mediated cytotoxicity, and these differences were statistically significant, suggesting that this sample collection was successful and can be used for subsequent analyses (Fig. [117]3). Fig. 3. [118]Fig. 3 [119]Open in a new tab Representative pathways enriched by GSEA. (A) Complement and coagulation cascades. (B) IL-17 signaling pathway. (C) Antigen processing and presentation. (D) Cell cycle. (E) Cytokine-cytokine receptor interaction. (F) TNF signaling pathway. (G) PI3K/Akt signaling pathway. (H) NK-mediated cytotoxicity. WGCNA This study aimed to identify the key genes involved in TIC in severe injury patients. To achieve this, the WGCNA algorithm was employed to construct a gene coexpression network. The gene expression profile specific to the TIC subtype of patients with severe injury was initially assessed using cluster analysis (Fig.[120]S1 A). Building a weighted gene network involves selecting the soft threshold power. Analysis of the pick soft threshold function revealed that with a soft threshold power of 4, the scale-free topology fit index curve flattens out at ~ 0.8, and the mean connectivity is relatively high, indicating sufficient information content (Fig.[121]S1 B). Therefore, a power of 6 was determined as the appropriate soft-thresholding value. After this transformation, the genes were clustered using the average linkage hierarchical clustering method. After merging the strongly associated modules using a 1.1 clustering height limit (Fig.[122]S1 C), 35 modules were identified for further study. In defining the gene network modules, this study adhered to a criterion requiring a minimum of 30 genes per module. This criterion was based on the standard set by the hybrid dynamic tree-cutting method. The dynamic tree-cutting method was utilized to identify distinct gene modules. Once these modules were established, the characteristic gene values (eigengenes) for each module were calculated to further understand their roles and interactions within the context of the TIC subtype of severe injury patients (Fig.[123]S1 D). To assess the correlation between each module and the disease, a heatmap depicting the module-trait relationships was generated using the Spearman correlation coefficient (Fig. [124]4A). Within the above modules, 3579 genes correlated with the phenotypes. The correlation between the modules was examined, and results showed no significant association between them (Fig. [125]4B). The reliability of the module delineation was further demonstrated by the transcription correlation analysis within the modules, revealing no substantial linkage between the modules (Fig. [126]4C). To visualize the correlation between each clinical phenotype and module, the linkET package was utilized to generate a correlation heatmap, which provided a clear display of the association between EPL, LY30, TT, R, C1_value, and each module (Fig. [127]4D). The “brown” “dark-gray”, “dark-red” and “light-yellow” modules, which are associated with the clinical variable phenotype of interest, revealed a strong association between the TIC and control groups (AD: r = 0.29, p = 1e–06; MetS: r = 0.1, p = 0.08) (Fig. [128]4E). Fig. 4. [129]Fig. 4 [130]Open in a new tab WGCNA revealing the potential genes of TIC. (A) Heatmap illustrating the relationships between modules and clinical phenotype features. Values in the small cells of the heatmap represent multiple calculated correlation coefficients between the eigenvalues of each trait and each module along with the corresponding statistically significant p-values. The color intensity corresponds to the strength of the correlation, with darker red indicating a more positive correlation and darker green indicating a more negative correlation. (B) Heatmap of collinear relationships among feature genes within modules, with red indicating strong positive correlations and blue indicating strong negative correlations. (C) Heatmap plot of the network demonstrating increasing coexpression interconnectedness through a gradient of increasingly saturated yellow and red colors. (D) Network and heatmap providing a detailed view of the relationships between traits, offering insights into the structure and coexpression patterns within the dataset. (E) Module membership (MM) vs. gene significance (GS) scatter plot of the TIC. Scatterplot illustrating the relationship between MM in coagulation-associated modules and GS for severe injury patients with TIC. Functional analysis of degs, TIC-associated genes, and WGCNA critical module genes The study focused on the intersection of critical module genes, DEGs, and TIC-associated genes using a Venn diagram, which identified 81 overlapping genes (Fig. [131]5A). To further understand the biological functions of these interesting genes, a functional analysis was conducted. GO enrichment analysis revealed that the genes are involved in processes such as regulation of DNA-binding TF activity, negative regulation of the immune system process, and negative regulation of response to external stimulus in BP; collagen-containing extracellular matrix (ECM), blood microparticle, and vesicle lumen in CC; and enzyme inhibitor activity, growth factor binding, and peptidase regulator activity in MF (Fig. [132]5B). KEGG analysis showed that these genes are involved in fluid shear stress and atherosclerosis, mitogen-activated protein kinase signaling pathway, IL-17 signaling pathway, and especially complement and coagulation cascades, which we are interested in (Fig. [133]5C). In addition to the above analyses, a petal diagram was employed to provide a detailed display of the pathways of interest. As illustrated in the diagram, within “complement and coagulation cascades,” enrichment of specific genes, such as TFPI, PLAT, C1QC, and A2M, was observed. This visual representation highlighted the presence of these genes in the pathway, offering deeper insights into their functional roles (Fig. [134]5D). To better demonstrate the interaction relationships among the intersecting genes, these genes were uploaded to the STRING database, and Cytoscape was utilized for visual analysis. This approach allowed mapping and visually interpreting the complex network of gene interactions, enhancing the understanding of how these genes interconnect and their roles in the biological pathways under study (Fig. [135]5E). Fig. 5. [136]Fig. 5 [137]Open in a new tab Enrichment analysis of KEGG pathways and visualization of networks involving the overlapping genes. (A) A Venn diagram displaying the intersection of GENECARDs, DEGs, and WGNA, revealing that 81 genes are expressed in all three categories. (B) GO analysis of gene intersections, encompassing BP, CC, and MF. The y-axis represents various GO terms, the x-axis indicates the gene counts enriched in corresponding GO terms, and the size of the circles reflects the number of genes. The 10 most highly regulated BP, CC, and MF were demonstrated. (C) Thirty most highly upregulated and downregulated pathways associated with gene intersections (generated by DAVID). (D) Interconnectedness of various biological pathways and genes. The circular nodes represent individual genes, with their size correlating with the significance of their role in the pathways. Colored lines denote the interaction between genes and the biological pathways in which they are involved. (E) PPI network of the intersecting genes, depicting a gene coexpression network where each node represents a gene and edges indicate coexpression relationships. Identification of feature genes by ML algorithms To identify feature genes, this study integrated three advanced ML algorithms, each contributing to a comprehensive analysis. SVM-RFE was first utilized, effectively narrowing down the candidate genes (Fig. [138]6A and B). Next, LASSO regression analysis was applied to select 13 predictive genes from statistically significant univariate variables (Fig. [139]6C and D). This method is renowned for its precision in high-dimensional data. Lastly, RF combined with feature selection was employed, analyzing the error rate and number of classification trees (Fig. [140]6E and F) to identify 46 genes of relative importance. To consolidate these findings, a Venn diagram was used, revealing four overlapping genes identified using all three methods (Fig. [141]6G). This multifaceted approach ensured the reliability and robustness of our gene selection process. The volcano plot was used in this research to further visualize the expression of feature genes. Through this graphical representation, the feature genes mostly exhibited an upregulated expression pattern (Fig. [142]6H). Fig. 6. [143]Fig. 6 [144]Open in a new tab Identification of central genes using ML techniques. (A and B) LASSO regression algorithm employed to refine and select pivotal gene variables, utilizing its capacity for regularization and dimensionality reduction to enhance model accuracy. (C and D) SVM-RFE algorithm systematically prioritizing genes by iteratively training the SVM model and eliminating features to isolate those with the most predictive power. (E and F) RF algorithm utilized for its ensemble learning approach, building multiple decision trees and merging their outcomes to identify genes most influential in classification accuracy. (G) Venn diagrams synthesizing the results from the three distinct algorithms, providing a visual intersection of hub genes consistently recognized across all methodologies, underscoring their potential significance as diagnostic markers. (H) Volcano plot used to visualize the genes selected through ML. Designing and evaluating a line graph for TIC diagnostic column data This study first employed a Venn diagram to illustrate the overlapping genes identified by SVM, LASSO, and RF algorithms and WGCNA, pinpointing the key feature genes for further analysis (Fig. [145]7A). TIC diagnostic models were developed using column line graphs for crucial genes (TFPI, MMP9, ABCG5, TPSAB1, TK1, IGKV3D.11, SAMSN1, TIMP3, and GZMB), and their predictive accuracy was assessed with calibration curves, revealing minimal deviation between predicted and actual TIC risks (Fig. [146]7B and C). The diagnostic importance of these genes was further validated through ROC analysis and 95% CI evaluation. Each gene showed substantial diagnostic accuracy, with AUC values and 95% CIs as follows: TFPI [0.840 (0.648–1)], MMP9 [0.840 (0.654–1)], ABCG5 [0.760 (0.575–0.945)], TPSAB1 [0.780 (0.577–0.983)], TK1 [0.900 (0.768–1)], IGKV3D.11 [0.710 (0.502–0.918)], SAMSN1 [0.840 (0.648–1)], TIMP3 [0.850 (0.670–1)], and GZMB [0.670 (0.420–0.920)] (Fig. [147]7D). These results highlighted their reliability as TIC diagnostic markers. In DCA, the column line graph model outperformed standard approaches, especially beneficial at high-risk thresholds (0–1), and showed greater clinical utility than individual gene curves (Fig. [148]7E). Compared to the control group in the dataset, these genes exhibited significantly higher expression levels in the TIC group (Fig. [149]7G). Fig. 7. [150]Fig. 7 [151]Open in a new tab Development and validation of a nomogram-based diagnostic model. (A) Venn diagram displaying the intersection of SVM, LASSO, RF, and WGCNA, revealing that nine genes are expressed in all three categories. (B) Nomogram model diagram utilizing the expression levels of diagnostic marker genes in the combined training dataset with feature-based analysis. (C) Calibration curves demonstrating the accuracy of the nomogram’s predictions by comparing predicted probabilities with observed outcomes. (D) ROC curves to evaluate gene signatures in predicting TIC. Each curve represents a different gene, with AUC and 95% CIs detailed in the legend. (E) DCA for gene signatures in predicting TIC, illustrating the standardized net benefit across a range of high-risk thresholds. Each line represents a gene, with the “All” line assuming all patients have TIC and “None” assuming no patients have TIC. The x-axis reflects the high-risk threshold, whereas the y-axis shows the standardized net benefit. (G) Expression of the nine diagnostically relevant genes in patients with severe injury between the TIC and control groups. Immunoinfiltration analysis of immune cells In the immunoinfiltration analysis, five distinct methods were employed: XCELL, QUANTISEQ, MCPCOUNTER, EPIC, and CIBERSORT (Fig.[152]S2). However, these findings consistently showed that only the CIBERSORT method proved highly effective in accurately identifying immunoinfiltration results. Utilizing the CIBERSORT method, a gene expression matrix was used to estimate the infiltration ratios of 22 different immune cells. For each of the 20 samples, a histogram was created to visually represent the composition of these immune cells. Each histogram used a range of colors to indicate the percentages of each type of immune cell, with the total summing to 1 for each sample (Fig. [153]8A). Box plots were used to compare differences in immunoinfiltration between TIC and control samples (Fig. [154]8B). These plots indicated a higher infiltration of CD4 naïve T cells, resting NK cells, M0 macrophages, and activated mast cells than the control. B-cell memory cells, T-cell follicular helper cells, activated NK cells, M1 macrophages, resting dendritic cells, and eosinophils demonstrated no infiltration results. Further analysis was conducted to assess the correlations between these types of immune cells in the TIC and control samples (Fig. [155]8C). The resulting correlation heatmap highlighted significant associations among various immune cells, including CD4 naïve T cells, resting NK cells, M0 macrophages, and activated mast cells. The network graph and heatmap offer a comprehensive overview of how immune cells may interact or influence one another and their potential regulation by or impact on specific genetic markers, such as TFPI and MMP9. This could have implications for understanding the underlying mechanisms of immunoresponses or the progression of TIC, particularly those impacting the immune system (Fig. [156]8D). Fig. 8. [157]Fig. 8 [158]Open in a new tab Assessment of immunoinfiltration and analysis of immune-related correlations. (A) Bar plot of the distribution of 22 different immune cells in TIC and control samples. Each column in the graph represents a sample. (B) Box plot of the expression of the relevant genes in the collected sample sets. *p < 0.05; **p < 0.01; ***p < 0.001, Wilcoxon rank-sum test. (C) A correlation heatmap displaying the relationships among immune cell infiltrates. Red, positive correlations; blue, negative correlations. The intensity of the color reflects the strength of the correlation. (D) Network diagram and heatmap correlation analysis between TFPI, MMP9, ABCG5, TPSAB1, TK1, IGKV3D.11, SAMSN1, TIMP3, and GZMB and immune cell types. Lines in the network indicate the correlation strength. The heatmap shows the correlation coefficients (blue for positive and red for negative). *p < 0.05; **p < 0.01; ***p < 0.001. The color legend for Pearson’s r and Mantel’s p aids in the data interpretation. Multifactor regulation network construction Leveraging miRNet and NetworkAnalyst databases, networks delineating the relationships between miRNAs, hub genes, and TFs were developed using Cytoscape. This network intricately connects 8 hub genes, 64 miRNAs, and 53 TFs, establishing a visual map of regulatory interactions that govern gene expression (Fig. [159]9). TF prediction was based on the JASPAR database. Fig. 9. [160]Fig. 9 [161]Open in a new tab miRNA-gene-TF regulatory network. A regulatory network mapping the interactions between miRNAs, diagnostic biomarker genes, and transcript factors. Circular nodes represent entities where green indicates miRNAs, purple indicates transcript factors, orange indicates genes with moderate interaction, and red indicates genes with high interaction centrality. The central large red node represents a gene with a particularly high degree of connectivity, suggesting a pivotal role in the regulatory network. Lines connecting nodes signify regulatory interactions, with the node size reflecting the extent of each gene’s regulatory influence. This network provides insights into complex regulatory mechanisms at the molecular level. GSEA The signature genes (TFPI, MMP9, ABCG5, TPSAB1, TK1, IGKV3D.11, SAMSN1, TIMP3, and GZMB) and corresponding expression data were entered into the GSEA platform, utilizing the curated gene sets c5.go. v7.4.symbols, and c2.cp.kegg.v7.4.symbols for comprehensive GO and signaling pathway analysis, respectively. The top 10 results from GO and KEGG analyses are presented illustratively (Figures [162]S3 and S4). RT-PCR validation of diagnostic hub genes To validate the expression of hub genes at the transcriptional and protein levels, qRT-PCR analyses were conducted on TIC and control whole-blood RNA samples from patients. Due to the lack of statistically significant IDE expression during the initial screening process and the absence of a complete transcript for IGKV3D.11, primer design and validation for these two genes were not performed in the RT-PCR experiment. The mRNA expression levels of TFPI, MMP9, ABCG5, TPSAB1, TK1, SAMSN1, and TIMP3 were significantly increased in the TIC groups compared to controls (p < 0.01) (Fig. [163]10). Fig. 10. [164]Fig. 10 [165]Open in a new tab Transcriptional expression of TFPI, MMP9, ABCG5, TPSAB1, TK1, SAMSN1, and TIMP3 in whole blood from TIC patients and control controls. Mean ± standard deviation. ***p < 0.001; **p < 0.01; *p < 0.05 against the control group. Discussion For decades, extensive research has recognized coagulation factor deficits as a critical issue in patients who have experienced severe trauma, with numerous studies documenting the phenomenon^[166]42–[167]45. This condition has been corroborated through the successful implementation of treatment protocols that involve removing clotting factors. Such treatments typically include the infusion of a higher ratio of FFP and the administration of prothrombin complex concentrates^[168]46,[169]47, which have shown efficacy in restoring coagulation homeostasis in trauma patients. Several theories have been proposed to explain this occurrence, with direct loss of blood, ongoing consumption of clotting factors at the site of injury, and dilution due to resuscitative efforts being the primary contributors^[170]48. Moreover, it is becoming increasingly evident that a reduction in the synthesis of coagulation factors, potentially due to hepatic dysfunction or systemic inflammatory responses, may play a significant role in TIC pathophysiology. These insights have sparked further investigation into the intricate mechanisms of coagulation, prompting the development of more targeted and effective therapeutic interventions to combat coagulation factor deficits in severely injured patients. Recent advances in molecular research and bioinformatics have significantly enhanced our understanding of gene function and disease progression. Through enrichment analysis, which examines MF, BP, and CC, thorough investigations can now be conducted into the impacts of gene variation and coexpression on protein functions and diseases. The development of WGCNA has become increasingly prevalent. This technique identifies clusters or modules of highly correlated genes and explores their associations with various diseases and related phenotypes. This integrative approach offers a more comprehensive view of the complex interplay between genetics and disease mechanisms^[171]25. Multiple studies have employed bioinformatics analysis to illuminate the role of key or “hub” genes and their underlying molecular mechanisms in patients with coagulopathy. These studies focused on understanding how these central genes influence the complex processes involved in coagulation disorders, shedding light on genetic factors and pathways that contribute to these conditions. This approach has been instrumental in unraveling the intricate biological networks and mechanisms in coagulopathy, paving the way for more targeted and effective treatments^[172]49,[173]50. This is the first study to apply WGCNA bioinformatics analysis methods and ML algorithms to explore severe injury patients with TIC to the best of the authors’ knowledge. This study is a pioneering effort to integrate WGCNA with ML to identify specific biomarkers that differentiate severely injured patients with TIC from a control group. This research began by identifying 1014 DEGs in the TIC group compared to the control group, including 711 upregulated and 303 downregulated genes. WGCNA was employed to pinpoint a gene module, specifically the ME blue module, which showed a strong positive correlation with severe injury patients with TIC, based on their gene coexpression patterns. This study identified three modules (dark gray, dark red, and brown) positively correlated with various TIC clinical phenotypes, including parameters, such as EPL, LY30, R time, TT, Lac, ALT, AST, and creatinine, among others. Given the goal of discovering potential diagnostic markers for severely injured TIC patients, the GENECARDs database was leveraged to identify 3171 TIC-associated genes. This comprehensive approach offers a new perspective in searching for biomarkers in TIC and paves the way for more targeted and effective treatments for this condition. This study integrated data from WGCNA, DEGs, and GENECARDs to identify genes linked to TIC. This analysis revealed 81 genes consistently expressed across these datasets. Further examination through GO and KEGG analyses of these genes was conducted. This analysis highlighted the following significant GO terms: negative regulation of the response to external stimuli in BP, blood microparticles in CC, and enzyme inhibitor activity in MF. The key KEGG pathways included the IL-17 signaling pathway, complement and coagulation cascades, and PI3K/Akt signaling pathway. These findings suggested that external stimuli significantly trigger coagulation cascades mediated by inflammation in patients with severe injuries. This study explored the efficacy of recombinant factor VIIα in damage control resuscitation for vascular trauma, focusing on vessel repair^[174]51. Another related study used fluorescent microarray analysis to investigate gene transcription in endothelial dysfunction, aiming to identify critical pathways^[175]52. The SVM-RFE algorithm of ML can remove redundant factors and retain only variables related to the outcome; it is widely used to rank features and choose significant ones for classification. Among the most effective feature selection methods, SVM-RFE has successfully identified hub genes for various diseases^[176]53,[177]54. In this study, 31 intersecting genes exhibited the highest accuracy and the lowest error in selecting candidate biomarkers. Using three ML methods (SVM-RFE, LASSO, and RF), nine hub genes (TFPI, MMP9, ABCG5, TPSAB1, TK1, IGKV3D.11, SAMSN1, TIMP3, and GZMB) were identified. A nomogram was constructed, and its predictive value was evaluated for TIC in patients with severe injury. TFPI codes for a Kunitz-type serine protease inhibitor involved in blood coagulation, specifically in the tissue factor-dependent pathway. It plays a crucial role in initiating coagulation by regulating the factor VIIα-TF complex, which activates further proteases, leading to fibrin clot formation. The gene product inhibits factor X and the VIIα-TF complex, forming an autoregulatory loop in the coagulation process. Its inhibition is correct hemostasis in hemophilia animal models^[178]55,[179]56. The MMP9 gene encodes a zinc-binding endopeptidase that plays a significant role in the degradation of ECM components, including collagen fibers. This activity leads to plaque destabilization, free radical generation, and proinflammatory cytokine activation. The gene product is thus crucial in processes that involve remodeling or breaking down the ECM, with implications in various physiological and pathological conditions^[180]57. ABCG5 and ABCG8 are members of the ABC transporter family and function as half-type ABC transporters. Recent studies have discovered a close association between ABCG5 and the transport of vitamin K, well known for its critical role in blood coagulation^[181]58. TPSAB1, encoding α-tryptase, is linked to hereditary α-tryptasemia, an autosomal dominant condition causing symptoms, such as skin flushing, dysautonomia, and gastrointestinal issues, in a subset of the population^[182]59. TK1 is an enzyme involved in the DNA salvage pathway, which is crucial for regenerating thymidine for DNA synthesis and repairing DNA damage. This process involves thymidine transfer across cell membranes and its conversion to its monophosphate form within the cell by TK1^[183]60. IGKV3D.11, also known as immunoglobulin-κ variable 3D-11, is part of the immunoglobulin gene family and encodes the V region of the variable domain of immunoglobulin light chains, which are crucial for antigen recognition in the immunoresponse. These immunoglobulins, produced by B lymphocytes, can be membrane-bound or secreted^[184]61,[185]62. SAMSN1, a signaling adaptor protein in hematopoietic tissues and immune cells, regulates inflammation and is upregulated in activated human B cells. It inhibits B-cell proliferation but encourages their differentiation into plasma cells^[186]63,[187]64. TIMP3 belongs to the family of tissue inhibitors of metalloproteinases, playing a pivotal role in preventing ECM breakdown by inhibiting matrix metalloproteinase activity^[188]65. GZMB, which is part of the granzyme serine protease family, is renowned for its proapoptotic role. It is the most extensively researched granzyme, particularly in health and disease contexts, including its role in mitigating cardiac ischemia-reperfusion injury^[189]66,[190]67. A refined nomogram integrating nine diagnostic markers was developed, each with high AUC values and accurate calibration, demonstrating superior accuracy and reliability in diagnosing TIC. The clinical application of this tool is anticipated to significantly aid in the early detection of TIC. Current research on coagulation diseases, particularly in the context of cancer, highlights a close relationship between the coagulation system and tumor progression. Studies indicated that a hypercoagulable state could enhance immunoinfiltration and tumor cell development, contributing to a poorer prognosis in cancer patients^[191]13. TIC necessitates prompt and increased transfusion and is associated with higher risks of organ failure and mortality. This outcome stems from the intrinsic interaction between coagulation and inflammation, leading to extensive inflammatory and immune repercussions. These effects further escalate the risk of complications, such as organ dysfunction and thromboembolic events, in the aftermath of early trauma^[192]68. Leukocyte-driven inflammation is a key factor in TIC, primarily through fibrinogen oxidation and breakdown, indicating a profound interaction between immunoresponses and coagulation mechanisms during traumatic events^[193]69. Thus, various immunoinfiltration analysis techniques were employed to study changes in immune cells in patients with severe trauma leading to TIC. This subsequent research was primarily informed by insights from liverwort analysis. This analysis indicated elevated CD4 naïve T cells, resting NK cells, M0 macrophages, and activated mast cells in TIC cases. The connection between the nine identified genes (TFPI, MMP9, ABCG5, TPSAB1, TK1, IGKV3D.11, SAMSN1, TIMP3, and GZMB) and their roles in the immune system was investigated. The nine hub genes demonstrated a positive correlation with immune-related cells. Unfortunately, among these nine genes, only TFPI has some evidence of traumatic coagulopathy^[194]70. To discover more valuable information, GSEA’s GO and KEGG analyses were conducted on these nine genes, hoping to find valuable evidence related to traumatic coagulopathy. Through GSEA’s GO and KEGG analyses, most of the nine genes were related to complement and coagulation cascades and closely related to lipid and fatty acid metabolism. Prospective epidemiological studies showed that dietary fatty acids can significantly influence atherosclerosis development and progression and myocardial infarction (MI) risk, suggesting a complex relationship between fatty acids and blood coagulation, as atherosclerosis and MI are closely related to coagulation processes^[195]71. Recent insights into nonalcoholic fatty liver disease pathophysiology suggested that it involves a complex pattern of prooxidative, proinflammatory, and prothrombotic components. These findings implied that excessive fatty acids are linked to coagulation dysfunctions, reinforcing the connection between lipid metabolism and coagulation processes^[196]72. An ongoing debate exists about the effect of high-dose ω−3 fatty acids on platelet aggregation and coagulation. Some studies suggested a positive impact on cardiovascular disease, whereas others showed no significant effect on bleeding time and platelet inhibition^[197]73. This provides a new perspective and direction for subsequent research. Future research should focus on the experimental validation of these findings, exploring their translational potential in diagnostic and therapeutic applications for TIC. Patients with elevated TFPI expression may have an increased risk of developing trauma-induced coagulopathy, which warrants further investigation as a key focus of our upcoming research. Conclusions This study provides a novel and comprehensive analysis of the molecular mechanisms underlying TIC, leveraging advanced methodologies to identify key genes, pathways, and networks. The findings offer significant implications for diagnosing and treating TIC, paving the way for future research in this critical area of trauma care. Electronic supplementary material Below is the link to the electronic supplementary material. [198]Supplementary Material 1^ (1.5MB, tiff) [199]Supplementary Material 2^ (2.2MB, tiff) [200]Supplementary Material 3^ (1.4MB, tif) [201]Supplementary Material 4^ (6.4MB, tiff) [202]Supplementary Material 5^ (16.9KB, docx) Acknowledgements