UCLA

We developed a web tool, SaVanTv2.0, that leverages cell type signatures, or genes highly expressed in cell types, for characterizing the underlying cell populations of gene expression data. SaVanTv2.0 offers an accessible web interface, diverse exploratory, inferential, and predictive analytics, and interpretable visualizations to distinguish cells across disease categories. We demonstrated SaVanTv2.0’s biomedical applications by (1) suggesting higher expression of neutrophils and lymphocytes and lower expression of monocytes in early-stage symptomatic compared with asymptomatic influenza patients, (2) distinguishing higher expression of B cells, monocytes, and neutrophils, and lower expression of natural killer cells and plasmacytoid dendritic cells in active compared with latent tuberculosis patients for a non-invasive diagnosis, and (3) contributing to the discovery of a T cell independent response that drives lepromatous leprosy. SaVanTv2.0 thereby carries out cell decomposition for the accessible and interpretable discovery of biomarkers and elucidation of immune responses in diseased samples.