February 23, 2023

Integrated Model of Single-cell Drug Sensitivity

1st Joint SDSC Call, Data-Intensive Research

Short Summary

The bone marrow of acute myeloid leukemia patients exhibits an extensive cellular heterogeneity with leukemic stem cells sitting at the apex of cancer cells hierarchy. These cancer stem cells are believed to be responsible for relapse and overall poor prognosis. Here we propose to develop a predictive modeling strategy to infer potential drug susceptibility of this highly relevant cell population. First, we will estimate cellular composition of patients by improving bulk deconvolution in the context of leukemia. We then propose to build a model of cell type drug sensitivity that will predict sensitivity of cancer stem cells to 122 drugs screened in the beatAML dataset. Informative features will be ranked and their functional relevance assessed. All together, we hope to pinpoint novel mechanisms of cancer stem cell sensitivity that will help to improve personalized therapeutical strategies in AML.

Goals

Building a model capable of disentangling the signal from drug screening assays conducted at the bulk level (comprising mixed cell populations) to elucidate the contribution of each distinct cell type.

Significance

Our model has the potential to assist pathologists in therapeutical decision-making. A more precise evaluation of the cancer subtype-specific drug sensitivity is crucial to offer each patient tailored treatments, reducing the need for aggressive therapies and enhancing survival rates.

Background

In 2018, Tyner and colleagues published the initial version of the beatAML cohort dataset, the most extensive repository of ex-vivo drug screening data across various cancer types. A year later, Van Galen and colleagues released a single-cell RNA-seq dataset providing comprehensive characterization of cancer cell populations in Acute Myeloid Leukemia patients with unparalleled detail.