Scholarlink

Background <p>Few cohorts have study populations large enough to conduct molecular analysis of <i>ex vivo</i> lung tissue for genomic analyses. Transcriptome imputation is a non-invasive alternative with many potential applications. We present a novel transcriptome-imputation method called the Lung Gene Expression and Network Imputation Engine (<i>LungGENIE</i>) that uses principal components from blood gene-expression levels in a linear regression model to predict lung tissue-specific gene-expression.</p> Methods <p>We use paired blood and lung RNA sequencing data from the Genotype-Tissue Expression (GTEx) project to train <i>LungGENIE</i> models. We replicate model performance in a unique dataset, where we generated RNA sequencing data from paired lung and blood samples available through the SUNY Upstate Biorepository (SUBR). We further demonstrate proof-of-concept application of <i>LungGENIE</i> models in an independent blood RNA sequencing data from the Genetic Epidemiology of COPD (COPDGene) study.</p> Results <p>We show that <i>LungGENIE</i> prediction accuracies have higher correlation to measured lung tissue expression compared to existing <i>cis-</i>expression quantitative trait loci-based methods (median Pearson’s <i>r</i> = 0.25, IQR 0.19–0.32), with close to half of the reliably predicted transcripts being replicated in the testing dataset. Finally, we demonstrate significant correlation of differential expression results in chronic obstructive pulmonary disease (COPD) from imputed lung tissue gene-expression and differential expression results experimentally determined from lung tissue.</p> Conclusion <p>Our results demonstrate that <i>LungGENIE</i> provides complementary results to existing expression quantitative trait loci-based methods and outperforms direct blood to lung results across internal cross-validation, external replication, and proof-of-concept in an independent dataset. Taken together, we establish <i>LungGENIE</i> as a tool with many potential applications in the study of lung diseases.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

LungGENIE: the lung gene-expression and network imputation engine

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年3月10日

查看全文
  • Auyon J. Ghosh,
  • Liam P. Coyne,
  • Sanchit Panda,
  • Aravind A. Menon,
  • Matthew Moll,
  • Michael A. Archer,
  • Jason Wallen,
  • Frank A. Middleton,
  • Craig P. Hersh,
  • Stephen J. Glatt,
  • Jonathan L. Hess

摘要

Background

Few cohorts have study populations large enough to conduct molecular analysis of ex vivo lung tissue for genomic analyses. Transcriptome imputation is a non-invasive alternative with many potential applications. We present a novel transcriptome-imputation method called the Lung Gene Expression and Network Imputation Engine (LungGENIE) that uses principal components from blood gene-expression levels in a linear regression model to predict lung tissue-specific gene-expression.

Methods

We use paired blood and lung RNA sequencing data from the Genotype-Tissue Expression (GTEx) project to train LungGENIE models. We replicate model performance in a unique dataset, where we generated RNA sequencing data from paired lung and blood samples available through the SUNY Upstate Biorepository (SUBR). We further demonstrate proof-of-concept application of LungGENIE models in an independent blood RNA sequencing data from the Genetic Epidemiology of COPD (COPDGene) study.

Results

We show that LungGENIE prediction accuracies have higher correlation to measured lung tissue expression compared to existing cis-expression quantitative trait loci-based methods (median Pearson’s r = 0.25, IQR 0.19–0.32), with close to half of the reliably predicted transcripts being replicated in the testing dataset. Finally, we demonstrate significant correlation of differential expression results in chronic obstructive pulmonary disease (COPD) from imputed lung tissue gene-expression and differential expression results experimentally determined from lung tissue.

Conclusion

Our results demonstrate that LungGENIE provides complementary results to existing expression quantitative trait loci-based methods and outperforms direct blood to lung results across internal cross-validation, external replication, and proof-of-concept in an independent dataset. Taken together, we establish LungGENIE as a tool with many potential applications in the study of lung diseases.