Constructing a 10‐core genes panel for diagnosis of pediatric sepsis

Abstract Background The lack of sensitivity and specificity of most biomarkers or the lack of relevant studies to demonstrate their effectiveness in sepsis. Methods Downloaded three sets of sepsis expression data (GSE13904, GSE25504, GSE26440) from GEO. Then, using the R limma package and WGCNA analysis tocore genes. Finally, the value of these core genes was confirmed by clinical samples. Results Compared to normal samples, we obtain many abnormally expressed genes in the pediatric sepsis. WGCNA co‐expression analysis showed that genes from blue and turquoise module were close correlation with pediatric sepsis. The top 20 genes (TIMP2, FLOT1, HCK, NCF4, SERPINA1, IL17RA, PGD, PRKCD, GLT1D1, ALOX5, SIRPA, DOK3, ITGAM, S100A11, ZNF438, PLIN3, LTB4R, TSPO, MAPK14, GAS7) of the blue module of pediatric sepsis were mainly enriched in neutrophil degranulation, etc The top 20 genes (TBC1D4, NOL11, NLRC3, ZNF121, DYRK2, ABCE1, MAGEH1, TMEM263, MCUB, MALT1, DDHD2, TRAC, NOC3L, LCK, TRMT61B, ZNF260, ENOPH1, LOC93622, NAE1, TRBC1) for turquoise module were mainly enriched in rRNA‐containing ribonucleoprotein complexes exported from the nucleus, etc The selected hub gene of pediatric sepsis was combined with the markers of cell surface and found 10 core genes (HCK, PRKCD, SIRPA, DOK3, ITGAM, LTB4R, MAPK14, MALT1, NLRC3, LCK). ROC showed that AUC of the 10 core genes for diagnosis of pediatric sepsis was above 0.9. Conclusion There were many abnormally expressed genes in patients with pediatric sepsis. The panel constructed by the 10 core genes was expected to become a biomarker panel for clinical application of pediatric sepsis.

pathophysiological mechanism of sepsis leads to the release of many biomarkers. By measuring more biomarkers, the host's response to infection can be better evaluated to better guide clinical treatment. 4 In the past two decades, many sepsis-related markers have been reported, and these biomarkers have been diagnosed due to a lack of consistent baseline in the study. The accuracy of these biomarkers was still unclear. 5 Many scholars were dedicated to the study of different types of biomarkers for early diagnosis of pediatric sepsis. And assess the role of prognosis. In general, biomarkers of pediatric sepsis can be classified according to their role as systemic inflammatory mediators 6 : ① molecules expressed on the phagocytic membrane [myeloid cell trigger receptor-1 (TREM-1), CD14 receptor body, CD163 receptor]. ② cytokines, such as interleukin (IL)-6, IL-8, IL-10. ③ acute phase proteins, such as procalcitonin (PCT) or C-reactive protein (CRP). However, the lack of sensitivity and specificity of most biomarkers or the lack of relevant studies demonstrates their effectiveness.
F I G U R E 1 Flow chart of the study Therefore, it was particularly urgent to look for more specific and more sensitive markers for predicting pediatric sepsis. We expect to obtain some genes as pediatric sepsis markers of diagnosis and prognosis.

| Study population
Three sets of pediatric sepsis gene expression data and their corresponding platform annotation data were downloaded from the

| Bioinformation analysis
The work flow chart was shown in Figure 1.

| Clinical patient verification
To analyze the clinical data of 48 patients with pediatric sepsis who were admitted to the Department of Newborn Critical Care Medicine (NICU) of the First Affiliated Hospital of Wenzhou Medical University from July 2010 to January 2019. In the study group, there were 48 cases, 23 males and 25 females; of which 15 were severely traumatized, 10 were abdominal injuries, 5 had larger burns, 10 were severe pneumonia, and 8 were acute peritonitis; aged 2-13 years old, mean age (5.8 ± 1.3) years old; the length of hospitalization was 10-28 days; the average length of hospitalization was (16.6 ± 5.2) days. In the control group, 50 cases, 24 males and 26 females, were normal children and examined in hospitals. There was no statistically significant difference in general data between the two groups (p > 0.05).

| qPCR detected the 10 core genes in sepsis
Extraction of serum RNA: Serum tRNA extraction kit (mirVanaPARIS Kit; AM1556) Ambion, USA. The main steps were as follows: Take 350 μl of serum and perform total RNA extraction according to the kit instructions. Total RNA was dissolved in 50 μl Ambion RNA eluate, and the final amount was about 45 μl. Serum total RNA quality testing primarily excludes genomic DNA contamination. The total RNA extracted from 350 μl of serum was approximately 500 ng, and any sample larger than 500 ng in total will be rejected. In order to eliminate genomic DNA contamination, gradient genomic DNA was added as a positive control in the qRT-PCR experiment, and the first strand was reverse-transcribed to synthesize cDNA. Design-specific primers according to the prime sequence of these genes (Table 1) optimize the optimal PCR conditions, use SYBR Green I RT-PCR to amplify the target fragment, and detect the fluorescence intensity of the product in real time. The internal reference uses the human β-actin gene. The relative content of the sample to be tested was calculated according to the 2 −ΔΔCT formula. 7 The study met the medical ethics standards and was approved by ethics committee of the First Affiliated Hospital of Wenzhou Medical University. All treatments and examinations were informed by the patient or family.

| Statistical analysis
Statistical analysis was performed by SPSS software v19 (SPSS Inc, USA). The data of each group were tested for normality and homogeneity of variance first. The data that accorded with normal distribution and homogeneity of variance were analyzed by one-way analysis of variance. The data that did not conform to normal distribution and homogeneity of variance were analyzed by nonparametric analysis. The normal distribution data were represented by "average standard ± deviation," and the non-normal distribution data were represented by the median (interquartile range). The receiver operating characteristic (ROC) curve was then used to estimate the classification performance.
Differences were considered statistically significant with p < 0.05.  used to express the spectral data, the probes were compared to the gene symbol using the platform annotation file, and the differentially expressed genes were screened using the Rlimma package.

| Differentially expressed genes in pediatric sepsis and normal group
According to the multiple of difference (|logFC| > 0.585) and significance threshold (FDR < 0.05), a total of 758 differentially expressed genes were screened using R package limma, of which 580 were up-regulated and 178 were down-regulated and shown in Table S1 and Figure S1.

| Module network construction and hub gene screening
The cytoscape was used to construct the network of key pediatric sepsis modules. Figure 6 was the network diagram of the pediatric sepsis

| Function and pathway enrichment analysis of key modules of pediatric sepsis
EnrichR was used to analyze the function and pathway enrichment of the medium top20 genes in two pediatric sepsis critical modules (blue, turquoise). The GO function enrichment results of the blue F I G U R E 8 Pediatric sepsis 5 core genes experimental verification interaction. It shown that HCK, PRKCD, SIRPA, DOK3, ITGAM genes were interacted with molecules related to cell surface, plasma membrane, nucleus, and cytoplasm from the Innate DB database module are shown in Figure S2A,B,C. The KEGG pathway enrichment results were shown in Figure S3. The GO function enrichment results of the turquoise module are shown in Figure S4A,B,C. The KEGG pathway enrichment results are shown in Figure S5.

| Verification of pediatric sepsis biomarkers
Using the Innate DB database, the selected pediatric sepsis hub gene was combined with the marker on the cell surface to verify, and the validated marker was selected as the characteristic gene (core genes) for constructing the panel. 8

| The result of clinical sample verification
According to ROC figure, the AUC of 10 core genes panel (0.936), sensitivity (96.6%), and specificity (58.7%) was in diagnosis of pediatric sepsis ( Figure 11 and Table 2). Our results showed that the clinical outcomes of pediatric sepsis patients with 10 core genes panel were consistent with bioinformatic predictions, suggesting that the 10 core genes panel was expected to be a good biomarker spectrum for the diagnosis of pediatric sepsis.

| D ISCUSS I ON
Sepsis is a common critical illness, and its high morbidity and mortality had attracted the attention of critical medical experts. The mortality rate of sepsis patients in this study was 26.11%, which was consistent with the mortality rate of sepsis in domestic and foreign literature. 9,10 In order to more effectively assess the prognosis of sepsis, scholars continue to explore more simple and effective monitoring indicators and methods. The APACHE II scoring system is one Taken together, it was difficult to achieve an accurate and comprehensive assessment of the prognostic indicators of a single sepsis, and recent studies had shown that multiple indicators need to be combined to evaluate. 12 Due to the lack of a genetic database for sepsis in adults, we tried to use the genetic database of pediatric sepsis to obtain some differential genes, and further verified it in clinical samples of adult sepsis.
Our findings showed that differential gene of pediatric sepsis was mainly enriched in neutrophil degranulation, neutrophil-mediated immunity, fibrin-1-rich particle cavity, cysteine, rRNA-contain- ROC analysis showed that AUC of the 10 core genes for diagnosis of pediatric sepsis was above 0.9. This result showed 10 core genes had a high diagnostic efficiency for pediatric sepsis.
In order to further verify the clinical value of 10 core genes, we found that the AUC of 10 core genes panel was 0.936 in diagnosis of pediatric sepsis. It helped clinicians to accurately identify the prognosis and diagnosis of pediatric sepsis in the early stage, optimize the management of pediatric sepsis treatment as soon as possible, and minimize mortality rate of the patients with pediatric sepsis.

| CON CLUS IONS
In conclusion, there were many abnormally expressed genes in patients with pediatric sepsis. The panel constructed by the 10 core genes was very effective in diagnosing of pediatric sepsis and was expected to become a biomarker panel for clinical application.

ACK N OWLED G M ENTS
This study was financially supported by the National Natural

CO N FLI C T O F I NTE R E S T S
The authors declare that they have no competing interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available in GEO database: GSE13904 (GPL570 platform), GSE25504 (GPL570, GPL6947, GPL13667 platform), GSE26440 (GPL570 platform) and these data were derived from the GEO database.