Th117 - Leveraging Single Cell Transcriptomics to Identify Pathogenic Cell Lineages

Zhaohui Du – Precision Medicine and Computational Biology – Sanofi US R&D; Samuel Lessard – Precision Medicine and Computational Biology – Sanofi US R&D; Tejaswi Iyyanki – Precision Medicine and Computational Biology – Sanofi US R&D; Dongyu Liu – Precision Medicine and Computational Biology – Sanofi US R&D; Clement Chatelain – Precision Medicine and Computational Biology – Sanofi US R&D; Franck Rapaport – Precision Medicine and Computational Biology – Sanofi US R&D; Shameer Khader – Precision Medicine and Computational Biology – Sanofi US R&D; Virginia Savova – Precision Medicine and Computational Biology – Sanofi US R&D

Abstract Text: The abnormal physiology of disease can be conceptualized as driven by a subset of cells either acquiring novel functions or losing canonical traits. While in cancer this is a well-established model for disease pathology, its application to autoimmune and neurological diseases is still under debate. In these diseases, a crucial missing step is identifying the lineages that drive disease, or “pathogenic” cell types. Here, we started from the basic assumption that genetic disease modifiers would have an increased impact on cell types that actively expressed the gene, and in turn that a pathogenic cell would be more likely to express genes that have a causal link to disease. In key autoimmune diseases, we combined single cell transcriptomics and human genetics to rank cell types by their likelihood of pathogenicity. We leveraged scRNA-seq data to identify cell type specific genes in each disease and intersected them with the disease’s casual gene modifiers. We found that the gene expression of genetic modifiers broadly clustered cells by their lineage, with several modifiers exclusively expressed in a single lineage. Moreover, we were able to rank cell types by their pathogenicity in atopic dermatitis, lupus nephritis, type 1 diabetes and inflammatory bowel syndrome, and validated the scoring by comparing it to an extensive literature review effort. The data-driven pathogenicity score was overall able to predict the pathogenic cell types cited in the literature and find new cell types potentially driving disease.