Deconvolution of Cell Type-Specific Drug Responses in Human Tumor Tissue with Single-Cell RNA-seq

Abstract

Precision oncology requires the timely selection of effective drugs for individual patients. An ideal platform would enable rapid screening of cell type-specific drug sensitivities directly in patient tumor tissue and reveal strategies to overcome intratumoral heterogeneity. Here we combine multiplexed drug perturbation in acute slice culture from freshly resected tumors with single-cell RNA sequencing (scRNA-seq) to profile transcriptome-wide drug responses. We applied this approach to glioblastoma (GBM) and demonstrated that acute slice cultures from individual patients recapitulate the cellular and molecular features of the originating tumor tissue. Detailed investigation of etoposide, a topoisomerase poison, and the histone deacetylase (HDAC) inhibitor panobinostat in acute slice cultures revealed cell type-specific responses across multiple patients, including unexpected effects on the immune microenvironment. We anticipate that this approach will facilitate rapid, personalized drug screening to identify effective therapies for solid tumors.Inter-and intra-tumoral heterogeneity present major challenges for cancer therapy. Precision medicine, or targeted therapy, entails the use of agents that preferentially target tumor cells based on unique molecular features. The success of this approach relies on extensive characterization of tumor heterogeneity and microenvironment. While scRNA-seq can determine the cellular composition of complex tumors and even reveal cell type-specific drug sensitivities, these measurements are ultimately limited by models of

Machine learning for perturbational single-cell omics

“…Factor decomposition methods break down expression values into different components of variation. Matrix factorization has proven popular, both in bulk (Squires et al, 2020;Zhao et al, 2020) and single cell (Mohammadi et al, 2020) genomics analyses, for its interpretability and scalability with larger datasets (Stein-O'Brien et al, 2018); for example, decomposing expression values in scRNA can cluster and rank genes into groups which are readily interpretable as mechanisms, pathways, or processes, most often through GO enrichment. Using this concept of factor decomposition, MUSIC (Duan et al, 2019) groups genes as ''topics'' through topic modeling and is able to quantify the size of the perturbational effect using the differential activation of all topics.…”

Integrating single-cell RNA-seq and imaging with SCOPE-seq2

“…The cell aneuploidy analysis was performed based on the scHPF model as described previously 23 . To compute the scHPF-imputed expression matrix, we multiplied the gene and cell weight matrix (expectation matrix of variable and in the scHPF model and then log-transformed the result matrix as .…”

Integrating single-cell RNA-seq and imaging with SCOPE-seq2

“…The cell aneuploidy analysis was performed based on the scHPF model as described previously (Zhao et al, 2020). To compute the scHPF-imputed expression matrix, we multiplied the gene and cell weight matrix (expectation matrix of variable θ and β) in the scHPF model and then log-transformed the result matrix as ( /10000 + 1) .…”