Identifying Key Cells for Fibrosis by Systematically Calling Cell Type-Phenotype Associations across Massive Heterogenous Datasets

Abstract

Fibrotic diseases pose a significant burden on health care, yet the key pathogenic fibroblasts involved remain unclear. We developed the fibrotic disease fibroblast atlas (FDFA), which comprises 394 single-cell and 38 spatial transcriptomic samples from 11 common fibrotic diseases. To perform a cell-type phenotype association study in large-scale heterogeneous datasets, we developed the single-cell phenotype association research kit for large-scale dataset exploration (SPARKLE). SPARKLE handles heterogeneity by incorporating confounding information into its generalized linear mixed models (GLMMs). The application of SPARKLE to FDFA revealed that matrix fibroblasts (MTFs) constitute a crucial pathogenic cell group in fibrosis. Their increased proportion correlate with the fibrotic process. MTFs also synergize with MYO-Fs, increasing their degree of fibrosis. Based on MTF, we identified 25 potential antifibrotic targets for broad-spectrum antifibrotic therapies. This study enhances our understanding of fibrosis and provides a reliable framework for large-scale cell type-phenotype association research. HighlightsO_LIA cross-tissue, multidisease fibrotic disease fibroblast atlas (FDFA) reveals diverse fibroblast subtypes in fibrosis diseases. C_LIO_LIA new toolkit, SPARKLE, is introduced for detecting robust cell type-phenotype associations in large-scale heterogeneous datasets. C_LIO_LIMultiple pathogenic fibroblasts associated with common fibrosis disease (MTF) onset and progression have been identified, suggesting potential cell-specific therapeutic strategies for fibrosis-related diseases. C_LI