Tumors are complex ecosystems composed of heterogeneous cell populations. Understanding the clonal cellular composition of the tumor and the non-malignant cells within the tumor ecosystem provides significant insights in the tumor recurrence, treatment, initiation, progression and metastasis. Previous studies estimated immune cell type content in bulk tumor expression data using immune cell signatures generated from peripheral blood mononuclear cells. However, with the advent of single-cell RNA sequencing methods, we can now also estimate the tumor associated non-malignant and malignant cell type contents. In this proposal, we describe a novel deep net approach for deconvolving different cell types in bulk tumor using single-cell sequencing data (scDEC). We will also infer tumor associated copy number variation (CNV) clones and their signatures from single-cell RNA sequencing data using our novel multiscale resolution signal processing based algorithm. Our approach will estimate not only the content of different immune cell types and tumor associated non-malignant cell types but also the content of different CNV clone types in bulk tumor. Moreover, we will discover new associations between cell type content and sample phenotype such as disease survival, subtype and outcome. Our proposed project will lead to major improvements in clinical care to guide the treatment and prognosis of various types of cancer.
Here we describe a novel multiscale resolution and deep net based approaches for deconvolving different cell types in bulk tumor using single cell RNA sequencing data (scDEC). Using our scDEC algorithm, we will identify novel associations between different cell type contents including immune cell, non-malignant tumor associated cell and CNV clone type contents and sample phenotypes such as disease survival. Our work will provide important prognostic insights and advancing therapeutic strategies that address the complexity of various cancer types.
