生物信息学联合机器学习探究主动脉瓣膜钙化潜在的诊断标志物

Potential diagnostic markers for aortic valve calcification：A study based on bioinformatics and machine learning

摘要

目的 通过机器学习方式从免疫、干细胞、成骨3个层面寻找主动脉瓣膜钙化潜在的诊断标志物。 方法 利用R语言将高通量基因表达数据库（Gene Expression Omnibus，GEO）中GSE51472、GSE12644和GSE55492 3个数据集的数据进行均一化后合并分析，得到差异表达基因。从Amigo数据库下载免疫、干细胞、成骨相关基因集分别与差异基因取交集，得到钙化瓣膜组织中异常表达的免疫、干细胞、成骨相关基因。利用机器学习方法对差异基因进行深度学习，寻找潜在的诊断生物标志物。进行基因富集（gene set enrichment analysis，GSEA）分析。临床标本结合细胞学实验验证诊断标志物的应用价值。 结果 3个数据集合并分析得到102个差异基因，与免疫相关的有51个，与干细胞相关的有1个，与成骨相关的有2个。GSEA分析显示无论是免疫相关基因还是干细胞、成骨相关基因均涉及免疫调控通路。多种机器学习合并分析结果显示白介素受体7（interleukin 7 receptor，IL7R）最具有诊断钙化性主动脉瓣病的价值，但临床标本并不支持生物信息分析结果。细胞实验显示IL7R可以促进瓣膜间质细胞增殖、成骨钙化。 结论 IL7R可能不是外周血诊断钙化性主动脉瓣疾病（calcific aortic valve disease，CAVD）的标志物，但是它在CAVD疾病演变过程中可能有重要作用，因此进一步探究IL7R在CAVD中的作用很有必要。

Abstract

Objective To identify potential diagnostic markers for aortic valve calcification using machine learning algorithms from the three aspects of immune response，stem cell activity，and osteogenesis. Methods R language was used for the homogenization，integration，and analysis of the data from GSE51472，GSE12644，and GSE55492 datasets in the GEO database to obtain differentially expressed genes. The gene sets associated with immune，stem cell，and osteogenesis were obtained from the Amigo database，which were intersected with the differentially expressed genes，respectively，to obtain the abnormally expressed genes associated with immune，stem cell，and osteogenesis in calcified valve tissue. Machine learning techniques were used to identify potential diagnostic biomarkers among the differentially expressed genes. In addition，Gene Set Enrichment Analysis（GSEA） was performed，and the diagnostic markers were validated using clinical specimens and cytology tests. Results A total of 102 differentially expressed genes were identified through the integration and analysis of the three datasets，among which there were 51 genes associated with immune response，1 gene associated with stem cells，and 2 genes associated with osteogenesis. The GSEA analysis showed that the genes associated with immune response，stem cells，and osteogenesis were all involved in immune regulatory pathways. The analysis based on multiple machine learning techniques showed that interleukin 7 receptor（IL7R） had the highest value in the diagnosis of calcific aortic valve disease （CAVD），but the results of bioinformatics analysis were not supported by clinical specimens. Cell experiments showed that IL7R could promote the proliferation of valvular interstitial cells and induce osteogenic calcification. Conclusion IL7R may not be a peripheral blood marker for diagnosing CAVD，but it may play an important role in the progression of CAVD，and therefore，it is necessary to further investigate the role of IL7R in CAVD.

Graphical abstract

关键词

钙化性主动脉瓣膜疾病 / 机器学习 / 生物标志物 / 生物信息

Key words

calcified aortic valve disease / machine learning / biomarkers / bioinformatics

基金资助

贵州省科技厅科技计划资助项目（编号：黔科合基础〔2020〕1Y424、黔科合基础-ZK〔2021〕一般396）