基于硬皮病线粒体相关基因的人工神经网络模型的构建

doi:10.12122/j.issn.1673-4254.2024.05.14

南方医科大学学报 ›› 2024, Vol. 44 ›› Issue (5): 920-929.doi: 10.12122/j.issn.1673-4254.2024.05.14

基于硬皮病线粒体相关基因的人工神经网络模型的构建

左志威¹(), 孟庆良¹, 崔家康¹, 郭克磊², 卞华¹^,²()

^1.河南中医药大学骨伤学院//河南省中医院风湿病科，河南郑州 450008
^2.南阳理工学院河南省张仲景方药与免疫调节重点实验室，河南南阳 473004

收稿日期:2023-11-20 出版日期:2024-05-20 发布日期:2024-06-06
通讯作者: 卞华 E-mail:15737264121@163.com;biancrown@163.com
作者简介:左志威，硕士，E-mail: 15737264121@163.com
基金资助:
国家自然科学基金(82074415);中原英才计划-中原科技创新领军人才项目(234200510006);河南省科技计划项目(232102311201)

An artificial neural network diagnostic model for scleroderma and immune cell infiltration analysis based on mitochondria-associated genes

Zhiwei ZUO¹(), Qingliang MENG¹, Jiakang CUI¹, Kelei GUO², Hua BIAN¹^,²()

^1.School of Orthopedics and Traumatology, Henan University of Traditional Chinese Medicine, Department of Rheumatology//Henan Provincial Hospital of Traditional Chinese Medicine, Zhengzhou 450008, China
^2.Henan Key Laboratory of Zhang Zhongjing Formulae and Herbs for Immunoregulation, Nanyang Institute of Technology, Nanyang 473004, China

Received:2023-11-20 Online:2024-05-20 Published:2024-06-06
Contact: Hua BIAN E-mail:15737264121@163.com;biancrown@163.com
Supported by:
National Natural Science Foundation of China(82074415)

摘要/Abstract

摘要：

目的建立基于GEO数据库硬皮病线粒体相关基因的机器学习和人工神经网络联合诊断模型并评价其效果。方法通过GEO数据库获取3份硬皮病芯片。其中GSE95065及GSE59785合并作为实验数据集并提取线粒体相关基因表达量，使用随机森林、LASSO回归和SVM算法筛选硬皮病线粒体相关特征基因，并用特征基因构建人工神经网络模型，用10折交叉验证模型准确性。来用验证数据集GSE76807对模型进一步验证，利用ROC曲线下面积值评估模型准确性。用RT-qPCR实验验证关键基因mRNA相对表达量。最后用CIBERSORT算法预估硬皮病与筛选出的潜在生物标志物的生物信息学关联。结果共获取差异基因24个，其中上调基因11个，下调基因13个。通过3种机器学习算法筛选到最相关的7个线粒体相关特征基因（POLB、GSR、KRAS、NT5DC2、NOX4、IGF1、TGM2），并构建人工神经网络诊断模型。使用该模型绘制了实验组和验证组诊断的ROC曲线，AUC值为0.984。验证组AUC为0.740。10折交叉验证AUC平均值大于0.980。RT-qPCR结果显示，与对照组相比，硬皮病中POLB（P=0.004）、GSR（P=0.029）、KRAS（P=0.007）、NOX4（P=0.019）、IGF1（P=0.008）、TGM2（P<0.0001）表达量明显上调，而NT5DC2（P=0.001）表达量在硬皮病组中明显下调。免疫细胞浸润显示，特征基因与滤泡辅助T细胞、幼稚B细胞、静息树突状细胞、记忆激活CD4⁺T细胞、巨噬细胞M0、单核细胞、记忆静息CD4⁺T细胞和肥大细胞激活等相关。结论构建了硬皮病特征基因的人工神经网络诊断模型，为探索硬皮病发病机制提供了一个新视角。

关键词: 硬皮病, 线粒体, 人工神经网络, 免疫细胞浸润, 机器学习

Abstract:

Objective To establish a diagnostic model for scleroderma by combining machine learning and artificial neural network based on mitochondria-related genes. Methods The GSE95065 and GSE59785 datasets of scleroderma from GEO database were used for analyzing expressions of mitochondria-related genes, and the differential genes were identified by Random forest, LASSO regression and SVM algorithms. Based on these differential genes, an artificial neural network model was constructed, and its diagnostic accuracy was evaluated by 10-fold crossover verification and ROC curve analysis using the verification dataset GSE76807. The mRNA expressions of the key genes were verified by RT-qPCR in a mouse model of scleroderma. The CIBERSORT algorithm was used to estimate the bioinformatic association between scleroderma and the screened biomarkers. Results A total of 24 differential genes were obtained, including 11 up-regulated and 13 down-regulated genes. Seven most relevant mitochondria-related genes (POLB, GSR, KRAS, NT5DC2, NOX4, IGF1, and TGM2) were screened using 3 machine learning algorithms, and the artificial neural network diagnostic model was constructed. The model showed an area under the ROC curves of 0.984 for scleroderma diagnosis (0.740 for the verification dataset and 0.980 for cross-over validation). RT-qPCR detected significant up-regulation of POLB, GSR, KRAS, NOX4, IGF1 and TGM2 mRNAs and significant down-regulation of NT5DC2 in the mouse models of scleroderma. Immune cell infiltration analysis showed that the differential genes in scleroderma were associated with follicular helper T cells, immature B cells, resting dendritic cells, memory activated CD4⁺T cells, M0 macrophages, monocytes, resting memory CD4⁺T cells and mast cell activation. Conclusion The artificial neural network diagnostic model for scleroderma established in this study provides a new perspective for exploring the pathogenesis of scleroderma.

Key words: scleroderma, mitochondria, artificial neural network, immune cell infiltration, machine learning

左志威, 孟庆良, 崔家康, 郭克磊, 卞华. 基于硬皮病线粒体相关基因的人工神经网络模型的构建[J]. 南方医科大学学报, 2024, 44(5): 920-929.

Zhiwei ZUO, Qingliang MENG, Jiakang CUI, Kelei GUO, Hua BIAN. An artificial neural network diagnostic model for scleroderma and immune cell infiltration analysis based on mitochondria-associated genes[J]. Journal of Southern Medical University, 2024, 44(5): 920-929.

图/表 12

图1 研究流程图

Fig.1 Flow chart of the study design.

表1 GEO数据库芯片数据集

Tab.1 GEO database chip data set

DATA	Sample size	Normal sample	SSc sample	Organization type	Data type
GSE95065	33	15	18	Homo sapiens	Expression profiling by array
GSE59785	82	2	80	Homo sapiens	Expression profiling by array
GSE76807	15	5	10	Homo sapiens	Expression profiling by array

表2 目的基因引物序列

Tab.2 Primer sequence of the target genes

Primer	Sequence 5'-3'
POLB	F: CTTCACTGGGAGTGACATCTTT R: CAGCGACTCCAGTGACC
GSR	F: GAGCTCCAAGTGGTGACTTC R: CAGGCCCTTAGAATTTGGGT
KRAS	F: GTGGATGAGTATGACCCTACG R GACCTGCTGTGTCGAGAATATC
NT5DC2	F: ACGTCGTCATCGTCCAG R: TCTCTAGGCGAGTGATACGG
NOX4	F: AGACTCTACACATCACATGTGG R: AAAGTTGAGGGCATTCACCA
IGF1	F: CCCACTGAAGCCTACAAA R: TTTCTTGTTTGTCGATAGGGA
TGM2	F: TGTCTGACAATGTGGAGGAG R: GCTGTAGCGAGAGGACATT
β-actin	F: TGCTGTCCCTGTATGCCTCTG R: TGATGTCACGCACGATTTCC

图2 硬皮病中线粒体相关差异基因表达分析

Fig.2 Analysis of the differential expressions of the differential mitochondria-related gene in scleroderma. A: DEGs heatmap (red for up-regulated and blue for down-regulated genes). B: DEGs volcano map (red for up-regulated and blue for down-regulated genes).

图3 差异基因的Metascape分析

Fig.3 Metascape analysis of the DEGs in scleroderma. A: DEG enrichment pathway and process network. B: Histogram of DEGs enrichment pathways and processes.

图4 差异基因的GO和KEGG富集分析

Fig.4 GO and KEGG enrichment analysis of DEGs in scleroderma. A: GO bar chart. B: GO circle. C: KEGG bubble diagram. D: KEGG cluster graph.

图5 机器学习算法筛选出的关键基因结果

Fig.5 Selection of the key genes using 3 machine learning algorithms. A: Correlation between the number of random forest trees and the model error. B: Result of Gini coefficient method in random forest classifier. C: Characteristic genes selected by LASSO regression algorithm. D: Feature genes screened by SVM algorithm. E: Venn diagram of the intersected genes of the 3 algorithms.

图6 10折交叉验证结果

Fig.6 10-fold cross-validation result.

图7 人工神经网络模型的构建与验证

Fig.7 Construction and verification of artificial neural network model. A: ANN result visualization. B: ROC curves of mitochondria-associated genes in the training dataset. C: ROC curves for mitochondria-related genes in the verification dataset.

图8 关键基因mRNA相对表达量

Fig.8 Relative mRNA expressions of the key genes in mouse models of scleroderma. *P<0.05, **P<0.01, ***P<0.001.

图9 免疫细胞表达相对含量相关性分析

Fig.9 Correlation analysis of immune cell expression. Red indicates a positive correlation and blue indicates a negative correlation.

图10 基因潜在生物标志物与浸润性免疫细胞的相关性

Fig.10 Correlation of the genetic biomarkers with the infiltrating immune cells. A: GSR. B: IGF1. C: KRAS. D: NOX4. E: NT5DC2. F: POLB. G: TGM2. P<0.05 indicates a significant correlation between immune cells and genes.

参考文献 35

1	Orteu CH, Ong VH, Denton CP. Scleroderma mimics - Clinical features and management[J]. Best Pract Res Clin Rheumatol, 2020, 34(1): 101489. DOI: 10.1016/j.berh.2020.101489
2	Li SC. Scleroderma in children and adolescents: localized Scleroderma and systemic sclerosis[J]. Pediatr Clin North Am, 2018, 65(4): 757-81. DOI: 10.1016/j.pcl.2018.04.002
3	Tsou PS, Sawalha AH. Unfolding the pathogenesis of scleroderma through genomics and epigenomics[J]. J Autoimmun, 2017, 83: 73-94. DOI: 10.1016/j.jaut.2017.05.004
4	Ge YZ, Zhou L, Chen ZX, et al. Identification of differentially expressed genes, signaling pathways and immune infiltration in rheumatoid arthritis by integrated bioinformatics analysis[J]. Hereditas, 2021, 158(1): 5. DOI: 10.1186/s41065-020-00169-3
5	Swanson MB, Weidemann DK, Harshman LA. The impact of rural status on pediatric chronic kidney disease[J]. Pediatr Nephrol, 2024, 39(2): 435-46. DOI: 10.1007/s00467-023-06001-0
6	Spinelli JB, Haigis MC. The multifaceted contributions of mitochondria to cellular metabolism[J]. Nat Cell Biol, 2018, 20(7): 745-54. DOI: 10.1038/s41556-018-0124-1
7	Uzhachenko R, Shanker A, Yarbrough WG, et al. Mitochondria, calcium, and tumor suppressor Fus1: At the crossroad of cancer, inflammation, and autoimmunity[J]. Oncotarget, 2015, 6(25): 20754-72. DOI: 10.18632/oncotarget.4537
8	Henderson J, Duffy L, Stratton R, et al. Metabolic reprogramming of glycolysis and glutamine metabolism are key events in myofibroblast transition in systemic sclerosis pathogenesis[J]. J Cell Mol Med, 2020, 24(23): 14026-38. DOI: 10.1111/jcmm.16013
9	Zhou X, Trinh-Minh T, Tran-Manh C, et al. Impaired mitochondrial transcription factor A expression promotes mitochondrial damage to drive fibroblast activation and fibrosis in systemic sclerosis[J]. Arthritis Rheumatol, 2022, 74(5): 871-81. DOI: 10.1002/art.42033
10	Leek JT, Johnson WE, Parker HS, et al. The sva package for removing batch effects and other unwanted variation in high-throughput experiments[J]. Bioinformatics, 2012, 28(6): 882-3. DOI: 10.1093/bioinformatics/bts034
11	Chen BY, Li YX, Yan YP, et al. Construction and analysis of heart failure diagnosis model based on random forest and artificial neural network[J]. Medicine, 2022, 101(41): e31097. DOI: 10.1097/md.0000000000031097
12	Rongvaux A. Innate immunity and tolerance toward mitochondria[J]. Mitochondrion, 2018, 41: 14-20. DOI: 10.1016/j.mito.2017.10.007
13	Kaufman BA, Van Houten B. POLB: a new role of DNA polymerase beta in mitochondrial base excision repair[J]. DNA Repair, 2017, 60: A1-A5. DOI: 10.1016/j.dnarep.2017.11.002
14	Sykora P, Kanno S, Akbari M, et al. DNA polymerase beta participates in mitochondrial DNA repair[J]. Mol Cell Biol, 2017, 37(16): e00237-17. DOI: 10.1128/mcb.00237-17
15	Couto N, Wood J, Barber J. The role of glutathione reductase and related enzymes on cellular redox homoeostasis network[J]. Free Radic Biol Med, 2016, 95: 27-42. DOI: 10.1016/j.freeradbiomed.2016.02.028
16	Xia YC, Wang GH, Jiang ML, et al. A novel biological activity of the STAT3 inhibitor stattic in inhibiting glutathione reductase and suppressing the tumorigenicity of human cervical cancer cells via a ROS-dependent pathway[J]. Onco Targets Ther, 2021, 14: 4047-60. DOI: 10.2147/ott.s313507
17	Niihori T, Aoki Y, Narumi Y, et al. Germline KRAS and BRAF mutations in cardio-facio-cutaneous syndrome[J]. Nat Genet, 2006, 38(3): 294-6. DOI: 10.1038/ng1749
18	Schubbert S, Zenker M, Rowe SL, et al. Germline KRAS mutations cause Noonan syndrome[J]. Nat Genet, 2006, 38(3): 331-6. DOI: 10.1038/ng1748
19	Groesser L, Herschberger E, Ruetten A, et al. Postzygotic HRAS and KRAS mutations cause nevus sebaceous and Schimmelpenning syndrome[J]. Nat Genet, 2012, 44(7): 783-7. DOI: 10.1038/ng.2316
20	Aslam A, Salam A, Griffiths CE, et al. Naevus sebaceus: a mosaic RASopathy[J]. Clin Exp Dermatol, 2014, 39(1): 1-6. DOI: 10.1111/ced.12209
21	Li RQ, Liu RQ, Zheng SY, et al. Comprehensive analysis of prognostic value and immune infiltration of the NT5DC family in hepatocellular carcinoma[J]. J Oncol, 2022, 2022: 2607878. DOI: 10.1155/2022/2607878
22	Li KS, Zhu XD, Liu HD, et al. NT5DC2 promotes tumor cell proliferation by stabilizing EGFR in hepatocellular carcinoma[J]. Cell Death Dis, 2020, 11(5): 335. DOI: 10.1038/s41419-020-2549-2
23	Qiu LX, Gong GC, Wu WJ, et al. A novel prognostic signature for idiopathic pulmonary fibrosis based on five-immune-related genes[J]. Ann Transl Med, 2021, 9(20): 1570. DOI: 10.21037/atm-21-4545
24	Jiménez SA, Castro SV, Piera-Velázquez S. Role of growth factors in the pathogenesis of tissue fibrosis in systemic sclerosis[J]. Curr Rheumatol Rev, 2010, 6(4): 283-94. DOI: 10.2174/157339710793205611
25	Piera-Velazquez S, Makul A, Jiménez SA. Increased expression of NAPDH oxidase 4 in systemic sclerosis dermal fibroblasts: regulation by transforming growth factor Β[J]. Arthritis Rheumatol, 2015, 67(10): 2749-58. DOI: 10.1002/art.39242
26	Cho SY, Oh Y, Jeong EM, et al. Amplification of transglutaminase 2 enhances tumor-promoting inflammation in gastric cancers[J]. Exp Mol Med, 2020, 52(5): 854-64. DOI: 10.1038/s12276-020-0444-7
27	Wang K, Zu CH, Zhang Y, et al. Blocking TG2 attenuates bleomycin-induced pulmonary fibrosis in mice through inhibiting EMT[J]. Respir Physiol Neurobiol, 2020, 276: 103402. DOI: 10.1016/j.resp.2020.103402
28	Tabata K, Mikita N, Yasutake M, et al. Up-regulation of IGF-1, RANTES and VEGF in patients with anti-centromere antibody-positive early/mild systemic sclerosis[J]. Mod Rheumatol, 2021, 31(1): 171-6. DOI: 10.1080/14397595.2020.1726599
29	Hamaguchi Y, Fujimoto M, Matsushita T, et al. Elevated serum insulin-like growth factor (IGF-1) and IGF binding protein-3 levels in patients with systemic sclerosis: possible role in development of fibrosis[J]. J Rheumatol, 2008, 35(12): 2363-71. DOI: 10.3899/jrheum.080340
30	Jin W, Zheng Y, Zhu P. T cell abnormalities in systemic sclerosis[J]. Autoimmun Rev, 2022, 21(11): 103185. DOI: 10.1016/j.autrev.2022.103185
31	DJrMesquita, Cruvinel WM, Resende LS, et al. Follicular helper T cell in immunity and autoimmunity [J]. Braz J Med Biol Res, 2016, 49(5): e5209. DOI: 10.1590/1414-431x20165209
32	Jaguin M, Fardel O, Lecureur V. AhR-dependent secretion of PDGF-BB by human classically activated macrophages exposed to DEP extracts stimulates lung fibroblast proliferation[J]. Toxicol Appl Pharmacol, 2015, 285(3): 170-8. DOI: 10.1016/j.taap.2015.04.007
33	Cardamone C, Parente R, Feo GD, et al. Mast cells as effector cells of innate immunity and regulators of adaptive immunity[J]. Immunol Lett, 2016, 178: 10-4. DOI: 10.1016/j.imlet.2016.07.003
34	Lescoat A, Ballerie A, Jouneau S, et al. M1/M2 polarisation state of M-CSF blood-derived macrophages in systemic sclerosis[J]. Ann Rheum Dis, 2019, 78(11): e127. DOI: 10.1136/annrheumdis-2018-214333
35	Maehara T, Kaneko N, Perugino CA, et al. Cytotoxic CD4⁺ T lymphocytes may induce endothelial cell apoptosis in systemic sclerosis[J]. J Clin Invest, 2020, 130(5): 2451-64. DOI: 10.1172/jci131700

[1]	王梓凝, 杨明, 李双磊, 迟海涛, 王军惠, 肖苍松. 心肌梗死后心肌纤维化小鼠心肌线粒体功能和能量代谢重塑相关性的转录组学分析[J]. 南方医科大学学报, 2024, 44(4): 666-674.
[2]	何慧珊, 郭二嘉, 蒙文仪, 王彧, 王雯, 何文乐, 吴元魁, 阳维. 基于磁共振图像机器学习放射组学模型预测脑胶质瘤的强化[J]. 南方医科大学学报, 2024, 44(1): 194-200.
[3]	叶红伟, 张钰明, 云琦, 杜若丽, 李璐, 李玉萍, 高琴. 白藜芦醇可减轻高糖诱导的心肌细胞肥大：基于促进SIRT1表达维持线粒体稳态[J]. 南方医科大学学报, 2024, 44(1): 45-51.
[4]	于佳池, 李芮冰, 夏天, 王佳楠, 金家丞, 袁漫秋, 李绵洋. 沉默PDCD4表达可减轻脓毒症血管内皮细胞损伤：基于改善线粒体动力学[J]. 南方医科大学学报, 2024, 44(1): 25-35.
[5]	王炼, 夏勇生, 张震, 刘馨悦, 施金冉, 王月月, 李静, 张小凤, 耿志军, 宋雪, 左芦根. 高表达MRPL13促进胃癌细胞增殖并影响患者预后：基于抑制p53信号[J]. 南方医科大学学报, 2023, 43(9): 1558-1566.
[6]	黄奕, 林丽珊, 黄浩华, 董航明. VDAC1通过诱导气道上皮细胞铁死亡参与屋尘螨诱导的哮喘小鼠气道炎症[J]. 南方医科大学学报, 2023, 43(8): 1333-1338.
[7]	王丽娅, 田美惠, 李蓉, 吴越, 王莎莎, 吕恒, 刘忠义, 于影. 乙醛脱氢酶2改善急性肺损伤小鼠的肺内皮屏障及维持线粒体动力学平衡[J]. 南方医科大学学报, 2023, 43(8): 1388-1395.
[8]	颜畅, 刘爽, 宋庆志, 胡艺冰. 二甲双胍通过抑制线粒体氧化磷酸化降低结直肠癌干细胞的自我更新能力[J]. 南方医科大学学报, 2023, 43(8): 1279-1286.
[9]	罗枭, 程义, 吴骋, 贺佳. 预测重症缺血性脑卒中死亡风险的模型：基于内在可解释性机器学习方法[J]. 南方医科大学学报, 2023, 43(7): 1241-1247.
[10]	宋陈芳, 黄镇河, 陈维, 王芳, 蔡梁凌, 赵斐, 赵悦. 恩格列净诱导的线粒体稳态显著减轻小鼠心脏微血管缺血/再灌注损伤[J]. 南方医科大学学报, 2023, 43(7): 1136-1144.
[11]	高凯绩, 王一豪, 曹海坤, 贾建光. 机器学习模型和Cox回归模型预测食管胃结合部腺癌预后的效能[J]. 南方医科大学学报, 2023, 43(6): 952-963.
[12]	徐文琴, 叶静静, 王飞, 陈天兵. 吡罗克酮乙醇胺盐通过PI3K/AKT通路破坏胶质瘤细胞的线粒体动力学[J]. 南方医科大学学报, 2023, 43(5): 764-771.
[13]	吴佳明, 邓忠权, 朱奕, 窦广健, 李进, 黄立勇. MicroRNA-431-5p在胃癌组织中低表达：基于线粒体和Bax/Bcl-2/caspase3信号通路[J]. 南方医科大学学报, 2023, 43(4): 537-543.
[14]	万璐, 钱宇池, 倪文静, 卢宇欣, 李巍, 潘艳, 陈卫东. 利格列汀通过激活AMPK/PGC-1α/TFAM通路改善糖尿病肾脏疾病线粒体生物合成[J]. 南方医科大学学报, 2023, 43(12): 2053-2060.
[15]	高毅男, 王培君, 逯素梅, 马万山. 甲基转移酶样3抑制剂STM2457通过调节线粒体功能改善代谢相关脂肪性肝病[J]. 南方医科大学学报, 2023, 43(10): 1689-1696.

基于硬皮病线粒体相关基因的人工神经网络模型的构建

An artificial neural network diagnostic model for scleroderma and immune cell infiltration analysis based on mitochondria-associated genes

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 12

参考文献 35

相关文章 15

编辑推荐

Metrics

本文评价