uniLIVER: a human liver cell atlas for data-driven cellular state mapping

Area of Science:

• Cellular and Molecular Biology
• Bioinformatics
• Genomics

Background:

• The liver's complex cellular composition is crucial for its metabolic and detoxification functions.
• Single-cell RNA sequencing (sc/snRNA-seq) offers unprecedented resolution for mapping tissue atlases.
• Understanding cellular heterogeneity is key to deciphering liver physiology and pathology.

Purpose of the Study:

• To construct a unified, high-resolution reference atlas of the normal human liver.
• To develop a novel computational method (LiverCT) for analyzing cellular states in query datasets.
• To investigate the prognostic implications of cellular states and zonation in liver cancer.

Main Methods:

• Integrative analysis of a large-scale dataset comprising 331,125 cells from 6 sc/snRNA-seq studies of normal human liver.
• Development and application of LiverCT, a machine learning approach for mapping query data to the reference atlas.
• Analysis of liver cancer datasets to identify "variant" cellular states and their correlation with clinical outcomes.

Main Results:

• Creation of uniLIVER, a comprehensive single-cell reference map of the human liver.
• Identification of "deviated" T cell states associated with stress pathways in hepatocellular carcinoma.
• Correlation of hepatocyte-cholangiocyte intermediate states and tumor cell zonation tendencies with poor prognosis.

Conclusions:

• The uniLIVER atlas provides a valuable resource for liver research.
• LiverCT enables the identification of clinically relevant cellular states and deviations from normal tissue.
• This framework offers a scalable approach for creating tissue-specific atlases and analyzing disease states.