Single-cell RNA sequencing and machine learning provide candidate drugs against drug-tolerant persister cells in colorectal cancer

Abstract

Drug resistance often stems from drug-tolerant persister (DTP) cells in cancer. These cells arise from various lineages and exhibit complex dynamics. However, effectively targeting DTP cells remains challenging. We used single-cell RNA sequencing (scRNA-Seq) data and machine learning (ML) models to identify DTP cells in patient-derived organoids (PDOs) and computationally screened candidate drugs targeting these cells in familial adenomatous polyposis (FAP), associated with a high risk of colorectal cancer. Three PDOs (benign and malignant tumor organoids and a normal organoid) were evaluated using scRNA-Seq. ML models constructed based on public scRNA-Seq data classified DTP versus non-DTP cells. Candidate drugs for DTP cells in a malignant tumor organoid were identified from public drug sensitivity data. From FAP scRNA-Seq data, a specific TC1 cell cluster in tumor organoids was identified. The ML model identified up to 36 % of TC1 cells as DTP cells, a higher proportion than those for other clusters. A viability assay using a malignant tumor organoid demonstrated that YM-155 and THZ2 exert synergistic effects with trametinib. The constructed ML model is effective for DTP cell identification based on scRNA-Seq data for FAP and provides candidate treatments. This approach may improve DTP cell targeting in the treatment of colorectal and other cancers. • scRNA-Seq and machine learning were used to identify drug-tolerant persister cells. • DTP cells were detected in organoids derived from familial adenomatous polyposis. • YM-155 and THZ2 were predicted to exert synergistic effects with trametinib. • ML methods can be used to identify drug-resistant cells in colorectal cancer.

Topics & Concepts