LiRNA: An interactive atlas of human liver RNAseq databases.

BACKGROUND & AIMS: While mouse models remain a cornerstone of mechanistic liver disease research, the translational relevance of mouse findings to human disease is frequently debated. To bridge this gap, studies increasingly attempt to validate mouse findings using human transcriptomic data. However, publicly available human liver RNA-sequencing (RNA-seq) databases are fragmented, inconsistent in their clinical phenotyping, and computationally inaccessible to many investigators. We sought to generate an interactive atlas of human liver RNA-seq datasets - LiRNA - to overcome these limitations.

METHODS

We identified 17 RNA-seq datasets generated from over 3,000 human liver biopsies and integrated them into a unified, open-access interactive web application (LiRNA) following harmonized FASTQ processing. We inferred biological sex from transcriptomic markers and genotyped four MASLD-associated variants (PNPLA3 rs738409, GCKR rs1260326, TM6SF2 rs58542926, MTARC1 rs2642438). We then systematically evaluated the generalizability of 64 previously reported gene-phenotype and gene-gene correlation findings from eleven leading hepatology and general science journals.

RESULTS

LiRNA accurately recapitulates established hepatic biology, capturing known fibrosis-associated transcripts, sexually dimorphic gene expression, HCV-response signatures, and genotype-specific transcriptional signatures across multiple datasets. When 64 recent translational findings were assessed against independent, larger cohorts in LiRNA, fewer than half generalized consistently.

CONCLUSIONS

LiRNA is an open-access, interactive platform integrating harmonized RNA-sequencing data from over 3,000 human liver biopsies. It facilitates contextualizing translational findings across diverse human populations, disease contexts, and clinical settings.

IMPACT AND IMPLICATIONS: Researchers increasingly attempt to validate mouse findings using human transcriptomic data. However, publicly available human liver RNA-sequencing (RNA-seq) databases are fragmented, inconsistent in their clinical phenotyping, and computationally inaccessible to many investigators. By lowering the computational barrier to multi-dataset analysis, LiRNA provides a resource for improving the rigor and generalizability of translational findings in liver disease research.