D109-16 AI-Powered Spatial Multi-Omics Reveals Vascular Endothelial Heterogeneity, Plasticity and Anti-Pulmonary Fibrosis Targets

Rationale

Capillaries are critical structures and functional components in the lung that facilitate gas exchange and maintain the stability of the gas-blood barrier. They exhibit spatial heterogeneity in the progression of pulmonary fibrosis, including unique functional subgroups, metabolic heterogeneity, and spatial distribution differences. Despite the development of anti-fibrotic drugs targeting capillaries, the therapeutic outcomes have been suboptimal. Therefore, identifying the pathogenic genes associated with capillary endothelial cells and clarifying their heterogeneity are essential for advancing treatment strategies.

Methods

We fine-tuned and perturbed the Geneformer model using a single-cell dataset of IPF (Idiopathic Pulmonary Fibrosis) from GSE136831 to identify pathogenic genes and therapeutic targets for capillary endothelial cells. Using FlowSig, we constructed the intercellular information flow between different cell types. Additionally, we employed Stereo-seq and MALDI MSI (Matrix-Assisted Laser Desorption Ionization Mass Spectrometry Imaging) for spatial transcriptomics and metabolomics sequencing on continuous tissue sections from IPF lung transplant samples. Using spatialMETA for multi-omics integration, we examined the spatial heterogeneity and metabolic characteristics of capillary endothelial cell distribution (Figure A).

Results

After fine-tuning Geneformer, the model accurately distinguished between different disease states (Figure B). Perturbation revealed 20 therapeutic pathogenic genes in capillary endothelial cells, with gene enrichment analysis showing significant associations with pathways related to pulmonary fibrosis and lipid metabolism (Figures C, D). Using FlowSig, we constructed an intercellular communication network across different IPF cell types (Figure E). Preliminary Stereo-seq analysis indicated that capillary endothelial cells exhibit extensive communication with other cell types in IPF (Figure F, G). Capillary endothelial cells influence myofibroblasts through Treg-cDC1-SMC signaling, distinguishing them from other endothelial cells (Figure H). MALDI MSI results showed distinct metabolic features in different regions of tissue sections (Figure I).

Conclusion

Through AI-based analysis of single-cell sequencing data, we identified 20 therapeutic pathogenic genes in capillary endothelial cells in IPF. Preliminary spatial multi-omics sequencing and analysis suggest that capillary endothelial cells influence myofibroblasts through Treg-cDC1-SMC signaling.

This abstract is funded by: Zhejiang University