Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2021 Mar 12:12:648329.
doi: 10.3389/fgene.2021.648329. eCollection 2021.

Deep Learning Reveals Key Immunosuppression Genes and Distinct Immunotypes in Periodontitis

Wanchen Ning  1 , Aneesha Acharya  2   3 , Zhengyang Sun  4 , Anthony Chukwunonso Ogbuehi  5 , Cong Li  6 , Shiting Hua  6 , Qianhua Ou  6 , Muhui Zeng  6 , Xiangqiong Liu  7 , Yupei Deng  7 , Rainer Haak  8 , Dirk Ziebolz  8 , Gerhard Schmalz  8 , George Pelekos  3 , Yang Wang  9 , Xianda Hu  7
Affiliations

Deep Learning Reveals Key Immunosuppression Genes and Distinct Immunotypes in Periodontitis

Wanchen Ning et al. Front Genet. .

Abstract

Background: Periodontitis is a chronic immuno-inflammatory disease characterized by inflammatory destruction of tooth-supporting tissues. Its pathogenesis involves a dysregulated local host immune response that is ineffective in combating microbial challenges. An integrated investigation of genes involved in mediating immune response suppression in periodontitis, based on multiple studies, can reveal genes pivotal to periodontitis pathogenesis. Here, we aimed to apply a deep learning (DL)-based autoencoder (AE) for predicting immunosuppression genes involved in periodontitis by integrating multiples omics datasets.

Methods: Two periodontitis-related GEO transcriptomic datasets (GSE16134 and GSE10334) and immunosuppression genes identified from DisGeNET and HisgAtlas were included. Immunosuppression genes related to periodontitis in GSE16134 were used as input to build an AE, to identify the top disease-representative immunosuppression gene features. Using K-means clustering and ANOVA, immune subtype labels were assigned to disease samples and a support vector machine (SVM) classifier was constructed. This classifier was applied to a validation set (Immunosuppression genes related to periodontitis in GSE10334) for predicting sample labels, evaluating the accuracy of the AE. In addition, differentially expressed genes (DEGs), signaling pathways, and transcription factors (TFs) involved in immunosuppression and periodontitis were determined with an array of bioinformatics analysis. Shared DEGs common to DEGs differentiating periodontitis from controls and those differentiating the immune subtypes were considered as the key immunosuppression genes in periodontitis.

Results: We produced representative molecular features and identified two immune subtypes in periodontitis using an AE. Two subtypes were also predicted in the validation set with the SVM classifier. Three "master" immunosuppression genes, PECAM1, FCGR3A, and FOS were identified as candidates pivotal to immunosuppressive mechanisms in periodontitis. Six transcription factors, NFKB1, FOS, JUN, HIF1A, STAT5B, and STAT4, were identified as central to the TFs-DEGs interaction network. The two immune subtypes were distinct in terms of their regulating pathways.

Conclusion: This study applied a DL-based AE for the first time to identify immune subtypes of periodontitis and pivotal immunosuppression genes that discriminated periodontitis from the healthy. Key signaling pathways and TF-_target DEGs that putatively mediate immune suppression in periodontitis were identified. PECAM1, FCGR3A, and FOS emerged as high-value biomarkers and candidate therapeutic _targets for periodontitis.

Keywords: autoencoder (AE); bioinformatics; deep learning; immunosuppression genes; periodontitis; therapeutic _targets.

PubMed Disclaimer

Conflict of interest statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Figures

FIGURE 1
FIGURE 1
Overall workflow. The flowchart depicts the autoencoder (AE) architecture and workflow combining deep learning (DL) techniques to identify key immunosuppression genes in periodontitis. Immunosuppression genes related to periodontitis from GSE16134 were applied as input features for an AE. The new transformed features in the bottleneck layer of the AE were clustered into different subtypes using K-mean clustering. Then, based on the clustering labels, we selected the top 100 most related genes from GSE16134 based on ANOVA F values. The input dataset was split at a 60%/40% ratio (training set/test set) to assess the robustness of the AE, using a 5-fold CV. Subsequently, based on the above labels of GSE16134, an SVM classifier was built and further applied for prediction in a validation set (GSE10334). To explore the biological roles of the different identified subtypes, differentially expressed genes (DEGs) and transcription factors (TFs), differential expression analysis, functional enrichment analysis, and construction of TF-_target DEGs interaction network were, respectively, applied. Eventually, to identify the immunosuppression genes that might be most pertinent to periodontitis, the overlapping DEGs among the DEGs discriminating disease (periodontitis) and controls and DEGs discriminating the subtypes classified with the AE and SVM models were determined.
FIGURE 2
FIGURE 2
Performance of the autoencoder (AE) and support vector machine (SVM) model. (A) Clustering results using the Silhouette index. Horizontal axis: Average silhouette width; Vertical axis: Number of clusters k. The optimal number of clusters is 2. (B) Clustering outcomes using Calinski–Harabasz criterion. Horizontal axis: Sum of the squared errors; Vertical axis: Number of clusters k. The optimal number of clusters is 2. (C,D) Comparison of AE with principal component analysis (PCA) based clustering. (C) The performance of AE based on Silhouette index. The optimum cluster number using AE is 2. Dim = dimensions. (D) The performance of PCA based on Silhouette index. The optimum cluster number using PCA is 6. Dim = dimensions. (E) Receiver operating characteristic (ROC) curve of the SVM model. Horizontal axis: false discovery rate (FDR); Vertical axis: true positive rate (TPR). The area under the curve (AUC) value of the GSE16134 test set is 97.72%.
FIGURE 3
FIGURE 3
Identification of the significant DEGs. (A) Intersection of DEGs discriminating sample type (disease vs. normal) (236 DEGs from GSE16134 and 194 DEGs from GSE10334) and DEGs of the disease samples classified into subtypes (subtype 1 vs. subtype 2) (219 DEGs from GSE16134 and 240 DEGs from GSE10334). (B,C) ROC curve of three significant genes (PECAM1, FCGR3A, and FOS) in GSE16134 (B) and GSE10334 (C). Horizontal axis: false discovery rate (FDR); Vertical axis: true positive rate (TPR).
FIGURE 4
FIGURE 4
The functional enrichment analysis of the overlapping DEGs common to the two datasets (GSE16134 and GSE10334). (A) The significantly enriched biological processes of the overlapped DEGs; (B) The significantly enriched signaling pathways of the overlapped DEGs.
FIGURE 5
FIGURE 5
Pathways enriched in the DEGs characterizing the two subtypes in GSE16134. (A) Top 20 enriched signaling pathways of DEGs in subtype 1. (B) Top 20 enriched signaling pathways of DEGs in subtype 2. (C) Heatmap shows the enriched signaling pathways of DEGs in the two subtypes.
FIGURE 6
FIGURE 6
Pathways enriched in the DEGs characterizing the two subtypes in GSE10334. (A) Top 20 enriched signaling pathways of DEGs in subtype1. (B) Top 20 enriched signaling pathways of DEGs in subtype 2. (C) Heatmap shows the enriched signaling pathways of DEGs in the two subtypes.
FIGURE 7
FIGURE 7
The transcription factor (TF)-_target interaction network of GSE16134 and GSE10334 involved in immunosuppression and periodontitis. Top 30 TFs were visualized in the network. Red and gray dots: up-regulated TF and DEG; Green and gray dots: down-regulated TF and DEG; Red dots: up-regulated DEG; Green dots: down-regulated DEG.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Alvarez C., Rojas C., Rojas L., Cafferata E. A., Monasterio G., Vernal R. (2018). Regulatory T lymphocytes in periodontitis: a translational view. Mediators Inflamm. 2018:7806912. 10.1155/2018/7806912 - DOI - PMC - PubMed
    1. Ambili R., Santhi W. S., Janam P., Nandakumar K., Pillai M. R. (2005). Expression of activated transcription factor nuclear factor-kappaB in periodontally diseased tissues. J. Periodontol. 76 1148–1153. - PubMed
    1. Andrukhov O., Hong J. S. A., Andrukhova O., Blufstein A., Moritz A., Rausch-Fan X. (2017). Response of human periodontal ligament stem cells to IFN-γ and TLR-agonists. Sci. Rep. 7:12856. 10.1038/s41598-017-12480-7 - DOI - PMC - PubMed
    1. Aoyagi T., Yamazaki K., Kabasawa-Katoh Y., Nakajima T., Yamashita N., Yoshie H., et al. (2000). Elevated CTLA-4 expression on CD4 T cells from periodontitis patients stimulated with Porphyromonas gingivalis outer membrane antigen: CTLA-4 expression in periodontitis. Clin. Exp. Immunol. 119 280–286. 10.1046/j.1365-2249.2000.01126.x - DOI - PMC - PubMed
    1. Bailly C. (2020). The implication of the PD-1/PD-L1 checkpoint in chronic periodontitis suggests novel therapeutic opportunities with natural products. Jpn. Dent. Sci Rev. 56 90–96. 10.1016/j.jdsr.2020.04.002 - DOI - PMC - PubMed

LinkOut - more resources

  NODES
innovation 2
twitter 2