Special Guest Seminar - Gene expression deconvolution of public cancer datasets

Gene expression microarrays are utilized extensively, especially in cancer research. However, most studies profile whole tissue samples that consist of a mix of cell types. A tumor tissue typically consists not only of tumor cells but also infiltrating immune cells and additional stromal cells. This greatly limits the conclusions derived from gene expression profiling of mixed cell samples. We developed a novel approach to blindly estimate the identity, relative proportions per sample and separated gene expression profiles of the cell types that constitute mixed cell samples. The only a-priori information needed is an initial estimate of the cell types in the tissue analyzed and their general reference signatures that may be easily obtained from public databases. This method is especially useful in re-analyzing the abundant datasets that exist in public repositories that are replete with gene expression microarrays of large patient pools from unique experimental conditions never to be repeated. However, the majority of these datasets are whole tissue, mixed cell samples for which additional information such as the identity, quantity or separated signatures of the constituent cell types rarely exist. We have applied our method to perform a comprehensive analysis of publicly available breast cancer datasets. Re-examination and analysis of cell type specific quantities and signatures holds great promise in discovering new phenomena otherwise not detected in the mixed cell samples of individual experiments.

Host: Prof. Ramit Mehr

--