Background: Molecular signatures are collections of genes characteristic of a particular cell type, tissue, disease, or perturbation. Signatures can also be used to interpret expression profiles generated from heterogeneous samples. Large collections of gene signatures have been previously developed and catalogued in the MSigDB database. In addition, several consortia and large-scale projects have systematically profiled broad collections of purified primary cells, molecular perturbations of cell types, and tissues from specific diseases, and the specificity and breadth of these datasets can be leveraged to create additional molecular signatures. However, to date there are few tools that allow the visualization of individual signatures across large numbers of expression profiles. Signature visualization of individual samples allows, for example, the identification of patient subcategories a priori on the basis of well-defined molecular signatures.
Result: Here, we generate and compile 10,985 signatures (636 newly-generated and 10,349 previously available from MSigDB) and provide a web-based Signature Visualization Tool (SaVanT; http://newpathways.mcdb.ucla.edu/savant ), to visualize these signatures in user-generated expression data. We show that using SaVanT, immune activation signatures can distinguish patients with different types of acute infections (influenza A and bacterial pneumonia). Furthermore, SaVanT is able to identify the prominent signatures within each patient group, and identify the primary cell types underlying different leukemias (acute myeloid and acute lymphoblastic) and skin disorders.
Conclusions: The development of SaVanT facilitates large-scale analysis of gene expression profiles on a patient-level basis to identify patient subphenotypes, or potential therapeutic target pathways.
Keywords: Heterogeneous samples; Molecular signatures; Tissue-specific expression; Transcriptomic analysis; Visualization tools.