LungGENIE: the lung gene-expression and network imputation engine

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Limited sample sizes hinder ex vivo lung tissue genomic analysis.
• Transcriptome imputation presents a non-invasive alternative for molecular analysis.
• Existing methods for gene expression prediction have limitations.

Purpose of the Study:

• To introduce Lung Gene Expression and Network Imputation Engine (LungGENIE), a novel transcriptome imputation method.
• To predict lung tissue-specific gene expression using blood gene-expression data.
• To provide a tool for genomic analysis in lung disease research.

Main Methods:

• Utilized paired blood and lung RNA sequencing data from the Genotype-Tissue Expression (GTEx) project for model training.
• Validated LungGENIE performance using unique paired blood and lung samples from SUNY Upstate Biorepository (SUBR).
• Demonstrated proof-of-concept in an independent dataset from the Genetic Epidemiology of COPD (COPDGene) study.

Main Results:

• LungGENIE achieved higher prediction accuracy (median Pearson's r = 0.25) compared to existing cis-expression quantitative trait loci (cis-eQTL) methods.
• Approximately half of the reliably predicted transcripts were replicated in the testing dataset.
• Showcased significant correlation between imputed and experimentally determined differential gene expression in COPD.

Conclusions:

• LungGENIE offers complementary and superior performance to existing cis-eQTL methods and direct blood-to-lung imputation.
• The tool demonstrated robust performance across internal validation, external replication, and independent datasets.
• LungGENIE is established as a valuable tool for studying lung diseases non-invasively.