Abstract

   Crohn’s disease (CD) is a recurrent, chronic inflammatory condition of
   the gastrointestinal tract which is a clinical subtype of inflammatory
   bowel disease for which timely and non-invasive diagnosis in children
   remains a challenge. A novel predictive risk signature for pediatric CD
   diagnosis was constructed from bioinformatics analysis of six mRNAs,
   adenomatosis polyposis downregulated 1 (APCDD1), complement component
   1r, mitogen-activated protein kinase kinase kinase kinase 5 (MAP3K5),
   lysophosphatidylcholine acyltransferase 1, sphingomyelin synthase 1 and
   transmembrane protein 184B, and validated using samples. Statistical
   evaluation was performed by support vector machine learning, weighted
   gene co-expression network analysis, differentially expressed genes and
   pathological assessment. Hematoxylin–eosin staining and
   immunohistochemistry results showed that APCDD1 was highly expressed in
   pediatric CD tissues. Evaluation by decision curve analysis and area
   under the curve indicated good predictive efficacy. Gene Ontology,
   Kyoto Encyclopedia of Genes and Genomes and gene set enrichment
   analysis confirmed the involvement of immune and cytokine signaling
   pathways. A predictive risk signature for pediatric CD is presented
   which represents a non-invasive supplementary tool for pediatric CD
   diagnosis.

   Keywords: pediatric Crohn’s disease, risk signature, diagnosis,
   prediction, immune

Graphical abstract

   graphic file with name j_biol-2022-0731-ga001.jpg
   [33]Open in a new tab

1. Introduction

   Crohn’s disease (CD) is a highly heterogeneous, recurrent chronic
   inflammatory condition of the gastrointestinal tract and is a clinical
   subtype of inflammatory bowel disease (IBD). The incidence of pediatric
   CD has shown a steep increase over the last 30 years [[34]1],
   particularly in China since 2000 [[35]2,[36]3], and one in four
   patients diagnosed with IBD was under the age of 18. Clinical
   manifestations of pediatric CD tend to be more severe and active than
   those of adult CD, and long-term consequences may include growth
   retardation [[37]4,[38]5]. Early diagnosis improves treatment outcomes,
   as cure rates and treatment tolerance are better in younger children
   [[39]6]. Thus, there is a need for additional diagnostic tools in the
   form of novel biomarkers to assist with timely and accurate diagnosis.
   Diagnoses depend on the assessment of clinical manifestations,
   laboratory tests, endoscopy, imaging and histopathology, but early
   clinical manifestations of nausea, vomiting, fever, abdominal pain and
   distention are nonspecific and atypical, making distinction from other
   abdominal diseases difficult [[40]7]. Pediatric CD also produces
   nonspecific laboratory, endoscopic, radiological and histological
   findings. Therefore, diagnostic tools using novel biomarkers are
   essential for the early detection of pediatric CD.

   Predictive risk signatures have been increasingly utilized for disease
   diagnosis, and pediatric CD pathogenesis is known to be related to
   individual gene profile variation [[41]8] and expression of risk genes,
   especially in younger children [[42]9,[43]10]. Geographical variation
   is also seen in genetic predisposition to pediatric CD, and common
   mutations of nucleotide oligomerization domain (NOD2)/CARD15 present in
   Caucasian patients have not been identified in Asian Han and Zhuang
   populations [[44]11,[45]12]. Susceptibility loci studies, including
   multi-ethnic, multi-center cohort and whole-genome sequencing analysis,
   have identified predictive variants for IBD [[46]13,[47]14]. Ethnicity
   appears to influence susceptibility site outcomes, and mutations of
   lipopolysaccharide-responsive beige-like anchor protein and cytotoxic T
   lymphocyte-associated protein 4 have been indicated as rare variants
   related to early-onset IBD, especially in African American populations
   [[48]14]. Some individual genetic markers of CD and IBD have been
   suggested for their diagnostic, prognostic and therapeutic value, but
   single genes give little information about the disease, and models
   composed of multiple genes may constitute a superior reflection of
   clinical features. However, little has been reported regarding
   predictive models for pediatric or adult CD.

   Machine learning (ML), including random forest (RF), support vector
   machine (SVM), K-nearest neighborhood, and Naïve Bayes approaches, have
   been employed in social sciences and medical research. ML algorithms
   are flexible and scalable, making them suitable for diagnosis and risk
   stratification [[49]15], with varying predictive power depending on the
   task and data type. A recent systematic review indicated the
   superiority of RF for disease prediction [[50]16], and the non-invasive
   nature of ML makes it a promising auxiliary diagnostic tool for IBD.

   A predictive risk signature was constructed from novel mRNA markers
   identified by weighted gene co-expression network analysis (WGCNA) and
   differentially expressed genes (DEGs) for pediatric CD diagnosis and
   validated by pathological assessment. The robustness of the current
   predictive risk signature and potential as a diagnostic tool for
   pediatric CD are demonstrated.

2. Materials and methods

2.1. Biopsy samples

   Pediatric CD patients were recruited through the Pediatric
   Gastroenterology service of Tongji Hospital during 2021–2022. Written
   informed consent was given by all participants, and ethical approval
   was granted by the Tongji Hospital Ethics Board (TJ-IRB20220756). All
   patients were under 17 and assigned to a CD group, diagnosed according
   to the modified Porto criteria [[51]17] with diagnosis confirmed by
   endoscopic biopsy or a control group of patients who were not diagnosed
   with CD. Tissue biopsies were collected by gastrointestinal endoscopy
   from gastrointestinal tract lesion sites and stained for pathological
   assessment. Detailed patient information is given in Table S1.

   Informed consent: Informed consent was obtained from all individuals
   included in this study.

   Ethical approval: The research related to human use has been complied
   with all the relevant national regulations and institutional policies
   and in accordance with the tenets of the Helsinki Declaration, and has
   been approved by the Tongji Hospital Ethics Board (TJ-IRB20220756).

2.2. Bioinformatics data information

   The gene expression profile of dataset [52]GSE10616, including 10
   ulcerative colitis (UC), 13 colon-only CD, 18 ileocolonic CD and 11
   normal samples, from the GEO online database
   ([53]https://www.ncbi.nlm.nih.gov/geo/) and annotated by the platform
   Affymetrix GeneChip Human Genome U133 Plus 2.0 Array. Clinical
   information related to the dataset is given in Table S2.

2.3. Data preprocessing and DEG analysis

   A total of 9,989 probes in the [54]GSE10616 dataset were normalized and
   DEGs between CD and normal samples were identified using “limma” R
   package in R.

2.4. Establishment of weighted co-expressed gene modules

   WGCNA is a well-established bioinformatics approach for building
   scale-free networks and conducting module analysis using gene
   expression profiles and was performed using the “WGCNA” and “pheatmap”
   R packages in R. Unavailable data and clinical subtypes of the dataset
   [55]GSE10616 were eliminated to ensure that gene expression profiles
   from pediatric CD and healthy control samples only were included. A
   scale-free network was formed, and the most suitable soft threshold (β
   = 13) was determined according to the computed average connection. A
   topological overlap matrix (TOM) and corresponding dissimilarity TOM
   (dissTOM) were computed, and hierarchical clustering, based on dissTOM,
   was carried out in the form of a systematic clustering dendrogram.
   mRNAs from [56]GSE10616 were categorized into 12 modules of 12
   different colors with a minimum module size of 100. The correlation
   between clinical traits and module eigengenes in each module was
   calculated.

2.5. ML model development and construction of a nomogram

   Two-stage selection was performed to identify the most relevant mRNAs
   for inclusion. Candidate mRNAs were selected from the intersection of
   DEGs and WGCNA hub genes and used to construct RF and SVM learning
   signatures. Five-fold cross-validation was used to verify the
   predictive performance of the signatures in the training cohort, and
   performance was evaluated by cumulative residual distribution analysis,
   boxplot distribution analysis and receiver operating characteristic
   (ROC) curve.

   A nomogram shows the occurrence probability of individual clinical
   diagnosis through aggregating scores of multiple predictor variables
   and is a frequently used tool used in the calculation of disease
   diagnosis risk. A nomogram to predict the risk of pediatric CD
   diagnosis was constructed based on the DEGs and WGCNA analysis by the
   “rms” R package. Calibration plots and area under the curve were also
   calculated to assess nomogram accuracy in predicting risk.

2.6. Pathway enrichment analysis

   The mRNAs in [57]GSE10616 were analyzed by gene set enrichment analysis
   (GSEA) 4.2.1 software Kyoto Encyclopedia of Genes and Genomes (KEGG)
   pathway analysis performed and visualized by “clusterProfiler,” the
   “pathview” and the “ggplot2” R package to identify pathways enriched in
   DEGs between control and CD samples. DEGs were assigned to pathways by
   gene ontology (GO) analysis, visualized by “clusterProfiler,” the
   “topGO” and the “ggplot2” R package.

2.7. Hematoxylin–eosin staining (H&E) and immunohistochemistry (IHC) assay

   Biopsy tissues were fixed with 4% paraformaldehyde for 48 h and covered
   with paraffin for sectioning into 4 μm sections and H&E staining
   (Servicebio, G1003, China) for histological analysis.

   IHC of colon sections from lesion sites was performed to investigate
   the expression of adenomatosis polyposis downregulated 1 (APCDD1).
   Colon tissues were immersed in 4% paraformaldehyde for 24 h, embedded
   in paraffin, and sectioned and antigen retrieval was performed by
   boiling slides for 20 min in ethylene diamine tetraacetic acid citrate
   buffer (pH = 6.0). Sections were incubated with anti-APCDD1 antibody
   (Signalway antibody, 43286, China, 1:50) at 4°C overnight, washed with
   PBS/Tween20 buffer (pH = 7.4) and anti-rabbit secondary antibody added
   for 50 min at room temperature. The 3,3′-diaminobenzidine chromogenic
   reagent kit was used to visualize staining under the microscope
   (SOPTOP, CX40, China), and samples were analyzed by ImageJ software
   (NIH, bundled with 64-bit Java 1.8.0_172).

2.8. Statistical analysis

   All data are expressed as mean  ±  SEM with Student’s t-test performed
   to determine statistical differences between groups. A value of p  < 
   0.05 was considered statistically significant.

3. Results

3.1. Identification of differentially expressed mRNAs in pediatric CD and
normal colon tissues

   A total of 561 DEGs with p <  0.05 and |log Fold Change| (|log FC|)  > 
   0.5 were identified from [58]GSE10616, including 376 upregulated and
   185 downregulated, and are shown in the volcano plot ([59]Figure 1a).

Figure 1.

   [60]Figure 1
   [61]Open in a new tab

   DEGs and WGCNA analysis of pediatric CD and control groups (a) Volcano
   plot for mRNAs based on p-value < 0.05 and |log FC| > 0.5. (b) An
   outlier was detected when the critical value was 400. (c) Sample tree
   and corresponding clinical traits. Clustering was visualized by Pearson
   correlation coefficient calculation. (d) Identification of the best
   soft threshold for WGCNA. The mean connectivity was the best with a
   soft threshold of 13. (e) Construction of connection distribution
   histogram and checking of range-free topology (β = 13). (f) Dendrogram
   of all genes clustered utilizing a dissimilarity measure (1-TOM). (g)
   Identification of 12 modules related to the clinical traits of
   pediatric CD.

3.2. Weighted co-expression network construction and clinically significant
module analysis

   The weighted co-expression network was constructed utilizing the gene
   expression matrix and clinical information in dataset [62]GSE10616
   through the R package of “WGCNA.” Data scrubbing of the expression
   profile was performed to exclude unavailable data and clinical
   subtypes, including “Internal control” and “Ulcerative colitis,” to
   ensure the inclusion of only suitable samples and genes. One outlier
   sample ([63]GSM267529) was screened out by Pearson correlation analysis
   with a height threshold of 400 ([64]Figure 1b) to ensure the stability
   of the co-expression network. Sample clustering with clinical traits
   was established ([65]Figure 1c), and the power of β = 13 was selected
   to be the soft-thresholding parameter to guarantee a scale-free network
   ([66]Figure 1d and e). A hierarchical clustering tree was constructed
   to complete the recognition of the co-expression network ([67]Figure
   1f). Twelve gene modules were identified with a minimum module size of
   100, and the correlation between clinical traits and modules was
   assessed by analyzing the eigengene of each module. The midnight blue
   module had the strongest correlation (r = 0.5, p = 7 × 10^−4) with
   clinical traits ([68]Figure 1g) and was selected for bioinformatics
   analysis of genes and DEGs.

   A total of 283 mRNAs with the highest connectivity (gene significance
   >0.2 and module eigengene-based connectivity [datKME] >0.8) were
   defined as hub mRNAs, considered to have the greatest degree of
   association with pediatric CD onset, and were recruited into the hub
   mRNAs of the midnight blue module (Table S3).

3.3. Construction of predictive risk signature

   The intersection of midnight blue hub mRNAs and DEGs, containing 155
   gene symbols ([69]Figure 2a). The intersection of the genes with p <
   0.05 and |logFC| > 0.1 and the midnight blue hub mRNAs contained 199
   mRNAs. By using these 199 mRNAs, a quantitative predictive signature
   assessment of pediatric CD risk. Six independent predictors were
   included APCDD1, complement component 1r (C1R), mitogen-activated
   protein kinase kinase kinase 5 (MAP3K5), lysophosphatidylcholine
   acyltransferase 1 (LPCAT1), sphingomyelin synthase 1 (SGMS1) and
   transmembrane protein 184B (TMEM184B). RF was shown to have the best
   performance through cumulative residual distribution analysis, boxplot
   distribution analysis ([70]Figure 2b) and ROC curve ([71]Figure 2c).
   Risk signature discrimination was calculated using a
   bootstrap-corrected C statistic, and the area under the ROC curve was
   0.991. The top 30 candidate mRNAs were identified from the RF model and
   gene significance plot ([72]Figure 2d and e) and a nomogram for
   quantitative predictive signature assessment of pediatric CD risk
   constructed from the six independent predictors by RF ([73]Figure 2f).
   The p values, logFC and average gene expression rates are shown in
   Table S4, and risk estimates are generated by aggregation of scores of
   each predictor in the nomogram. The closeness of the solid and dotted
   lines indicates the accuracy of the prediction, and the predicted risk
   was shown to match the incidence by the calibration curve ([74]Figure
   2g). A decision curve analysis (DCA) diagram was plotted to refine the
   predictive signature, and the predictive signature showed good net
   benefit across the threshold probability range: 0.2–0.9 ([75]Figure
   2h). The clinical impact curve was plotted using the signature
   ([76]Figure 2i), and good potential facilitation of clinical practice
   shown.

Figure 2.

   [77]Figure 2
   [78]Open in a new tab

   Establishment of a nomogram for risk prediction using the intersection
   between DEGs and the most clinically associated module in the WGCNA.
   (a) Intersection of hub mRNAs with DEGs, containing 155 mRNA symbols.
   (b) Cumulative residual distribution analysis and boxplot distribution
   analysis of RF and SVM. (c) ROC curve of RF and SVM. (d) Construction
   of RF model. (e) Gene significance of the top 30 candidate mRNAs. (f)
   Nomogram combining signatures with clinical traits. (g) Calibration
   curves comparing nomogram-predicted survival probabilities and measured
   survival probabilities of training and validation cohorts. y-axis:
   measured survival probabilities; x-axis: nomogram-predicted survival
   probabilities; dotted line: ideal prediction by an optimal signature;
   solid line: nomogram performance. (h) DCA of predicted nomogram
   signature. (i) Clinical impact curve.

   C1R, MAP3K5, LPCAT1 and SGMS1 have previously been linked to IBD, but
   any association of APCDD1 and TMEM184B with IBD has not been reported
   [[79]18,[80]19,[81]20,[82]21]. Differential expression of the six
   predictor mRNAs between pediatric CD and healthy intestinal tissue from
   dataset [83]GSE10616 was visualized in the volcano plot ([84]Figure
   1a).

3.4. Pathway enrichment analysis

   Pathway enrichment analysis was performed on the 199 intersecting mRNAs
   to associate affected pathways with pediatric CD. GO enrichment
   analysis indicated biological processes (BP) to include T-cell
   activation, leukocyte cell–cell adhesion, positive regulation of
   cytokine production and extracellular matrix organization; cellular
   components (CC), signaling pathways, such as collagen-containing
   extracellular matrix and azurophil granule ([85]Figure 3a) and
   molecular functions (MF), cytokine binding, immune receptor activity,
   growth factor activity, integrin binding and extracellular matrix
   constituent ([86]Figure 3a). These findings are consistent with the
   immune and autoimmune implications of IBD. Immune cell infiltration
   analysis of dataset [87]GSE10616 revealed increased infiltration of
   plasma cells and M1 macrophages (Figure S1). Increased collagen
   secretion and remodeling are pathophysiological components of the
   intestinal strictures and fistulas associated with CD, especially at
   the chronic stage [[88]22]. KEGG analysis of the pediatric CD group
   indicated enrichment of cytokines and associated pathways ([89]Figure
   3b and c).

Figure 3.

   [90]Figure 3
   [91]Open in a new tab

   GO and KEGG pathway enrichment analysis of the intersected genes in
   pediatric CD and control groups. (a) GO analysis of the intersected
   genes by BP, CC and MF. y-axis: GO terms; x-axis: gene ratio of each GO
   term. (b) KEGG pathway enrichment analysis of the intersected genes.
   y-axis: KEGG terms; x-axis: gene ratio of each KEGG term. (c) KEGG
   pathway enrichment analysis of the intersected genes. y-axis: KEGG
   terms; x-axis: gene counts of each KEGG term.

   GSEA showed pediatric CD samples to be enriched for pathways, such as
   cytokine–cytokine receptor interaction, chemokine signaling pathways,
   the intestinal immune network for IgA production and toll-like receptor
   signaling pathway ([92]Figure 4a–d), indicating the possibility that
   mycobiota orchestrate gut fungal commensalism by inducing IgA
   antibodies [[93]23]. Control samples were enriched for pathways related
   to fatty acid metabolism and steroid hormone biosynthesis ([94]Figure
   4e and f). Small intestinal epithelial cells from CD patients have
   previously shown evidence of impaired lipid peroxidation metabolism,
   and the abundant polyunsaturated fatty acid content of the Westernized
   diet may lead to enteritis [[95]24]. The top 20 enriched pathways and
   genes are shown in Table S5.

Figure 4.

   [96]Figure 4
   [97]Open in a new tab

   GSEA analysis of genes in pediatric CD and healthy controls from the
   [98]GSE10616 dataset. CD group genes were enriched in pathways
   including (a) cytokine-cytokine receptor interaction, (b) chemokine
   signaling pathways, (c) intestinal immune network for IgA production
   and (d) toll-like receptor signaling pathway. Healthy control group
   genes were enriched in pathways related to (e) fatty acid metabolism
   and (f) steroid hormone biosynthesis.

3.5. H&E staining, endoscopic evaluation and IHC validation of APCDD1
overexpression in pediatric CD lesion sites

   H&E staining of CD gastrointestinal biopsies showed damage, including
   shortening of the intestinal villi, inflammatory cell infiltration and
   lymphoid follicular formation, compared with controls ([99]Figure 5a).
   Multiple ulcer lesion sites could be seen during endoscopic examination
   of pediatric CD patients compared with normal mucosa in controls
   ([100]Figure 5b).

Figure 5.

   [101]Figure 5
   [102]Open in a new tab

   H&E staining, endoscopic views and IHC validation of APCDD1
   overexpression in pediatric CD lesion sites (a) H&E staining of control
   and CD samples (magnification ×400, scale bar = 25 μm). (b) Endoscopic
   views of control and CD groups. (c) Expression of APCDD1 in control and
   CD samples (n = 5, magnification ×400, scale bar = 25 μm). (d)
   Quantification of APCDD1 expression in healthy control and CD samples.
   Case1: terminal ileum; Case2: duodenal bulb; Case3: terminal ileum;
   Case4: descending colon; Case5: terminal ileum; Case6: descending
   colon. ***, p < 0.001.

   IHC staining showed significant overexpression of APCDD1 in CD lesion
   tissues compared with controls, consistent with the results of
   bioinformatics analyses ([103]Figure 5c and d).

4. Discussion

   The pediatric CD is an increasing global health problem with rising
   rates in China over the last 20 years due to risk factors, such as the
   adoption of a Westernized diet and industrial environments. Early
   diagnosis of pediatric CD is likely to shorten the course of therapy
   and improve prognosis, but existing diagnostic methods rely on judgment
   of symptoms, laboratory findings and imaging features. Patients have
   often entered the chronic and recurring inflammation stage prior to
   diagnosis to the detriment of treatment [[104]25], illustrating the
   need for complementary diagnostic tools and early screening tools.

   Causes of pediatric CD are multifactorial, including gene mutation,
   immunity and environmental factors, with the first two being considered
   more significant [[105]26]. Indeed, gene mutations and positive family
   history often lead to diagnosis at a younger age and a higher incidence
   of extra-intestinal manifestations.

   Previous studies of predictive risk signatures for pediatric IBD have
   focused on complications and drug reactions, and some have shown
   correlations with pathways or molecules. There are a few studies on
   risk loci for pediatric IBD, but none has previously established a
   predictive risk signature for the diagnosis of pediatric CD.

   The current study identified biomarkers from the gene expression
   profiles of dataset [106]GSE10616 in the GEO database to establish a
   predictive risk signature. The midnight blue module from WGCNA analysis
   had the lowest p-value, and intersection with DEGs allowed a nomogram
   based on six mRNAs, four of which have been previously linked to IBD,
   to be constructed. APCDD1 had the highest gene significance in the risk
   signature and is an inhibitor of the Wnt/β-catenin signaling pathway
   and the target of the β-catenin/T-cell factor 4 complex with probable
   involvement in colorectal carcinogenesis [[107]10]. The locus is a
   known variant in familial colorectal cancer patients [[108]27] but has
   not previously been associated with IBD. C1R encodes a proteolytic
   subunit in the complement system C1 complex of the innate immune
   response and is involved in CD progression. The C1R is bound by
   glucocorticoid receptors and is a potential serum biomarker of response
   to glucocorticoid treatment in pediatric IBD [[109]18]. LPCAT1 encodes
   an enzyme that converts lysophosphatidylcholine to phosphatidylcholine
   to regulate the number and size of lipid droplets and has been
   suggested as an indicator of UC remission [[110]28]. SGMS1 was a
   predicted target of baicalin in a study of the baicalin
   sphingolipid-linked treatment of colitis [[111]29]. TMEM184B encodes
   transmembrane protein 184B, a putative mitogen-activated protein kinase
   (MAPK)-activating protein, and was upregulated in antineutrophil
   cytoplasmic antibody-associated glomerulonephritis [[112]30]. MAP3K5 or
   ASK1 was shown to induce c-Jun N-terminal kinase 1 (JNK/MAPK8/SAPK1)
   phosphorylation, leading to Bcl-2 activation of cell autophagy, and was
   associated with endoplasmic reticulum stress, making it a promising
   therapeutic target for IBD [[113]20]. MAP3K5 is associated with
   autophagy-induced ER stress in IBD patients, allowing bacterial
   colonization of intestinal mucosa [[114]31]. The current study showed
   that these six predictor mRNAs were upregulated in CD samples and were
   associated with pediatric CD progression.

   A bioinformatics approach was used to intersect DEGs and genes with the
   highest clinical significance and module eigengene-based connectivity
   to determine a risk signature. The current novel bioinformatics study
   combined WGCNA, DEGs and risk signature, taking into account gene
   interaction network and expression profiles.

   We acknowledge some deficiencies in the present study. Publicly
   available gene profile data were used, and the sample size of the
   external clinical cohort was small. Only APCDD1 was verified in our
   predictive risk signature, and other predictor mRNAs require
   validation. In vivo and in vitro experiments are also required to
   supplement the IHC analysis, such as Western blotting and qPCR to
   assess protein expression and transcription rate of the other five
   predictors in HCT116 and HT29 cell lines.

5. Conclusion

   A predictive risk signature for the pediatric CD was constructed by
   bioinformatics methods based on six mRNAs to assist with the
   stratification and diagnosis of pediatric CD. Further investigation is
   required to combat pediatric CD.

Supplementary Material

   supplementary material

Footnotes

   Funding information: This work was supported by the Nature Science
   Foundation of Hubei Province (2020CFB685).

   Author contributions: Y.Z. and Q.J. performed the literature search,
   analyzed the data, conducted experiments and drafted the manuscript.
   T.Y.S.Y. and M.S. conducted the experiments and analyzed the data. Y.R.
   and S.L. edited and revised the manuscript and conceived the study. The
   authors applied the approach for the sequence of authors.

   Conflict of interest: Authors state no conflict of interest.

   Data availability statement: The datasets generated during and/or
   analyzed during the current study are available from the corresponding
   author on reasonable request.

Contributor Information

   Yuanyuan Zhan, Email: m202276253@hust.edu.cn.

   Quan Jin, Email: 344561172@qq.com.

   Tagwa Yousif Elsayed Yousif, Email: telsayed@jazanu.edu.sa.

   Mukesh Soni, Email: Mukesh.research24@gmail.com.

   Yuping Ren, Email: yprentj@163.com.

   Shengxuan Liu, Email: zooophelia170@163.com.

References