Abstract:
|
Tumor tissue samples comprise a mixture of cancerous and surrounding stromal cells. Understanding tumor heterogeneity is crucial to analyzing gene signatures associated with cancer prognosis and treatment decisions. Numerous computational approaches previously developed have limited ability to deconvolute heterogeneous tumor samples. We have developed a deconvolution tool, DeMix-T, that can explicitly account for a third component such as infiltrating immune cells, and address this challenging problem when the observed signals are assumed to come from a mixture of three cell compartments, infiltrating immune cells, the tumor microenvironment and tumor tissues. The optimization-based algorithm is computationally feasible when it needs to compute high-dimensional integrals. It therefore involves a novel two-stage filtering method that yields accurate estimates of cell proportions and compartment-specific expression profiles. Simulations and real data analyses have demonstrated the good performance of our method. DeMix-T allows for a further understanding of immune cell infiltration in cancer and assists in the development of novel prognostic markers and therapeutic strategies.
|