O03_04

Application of machine learning to single-cell RNA sequencing provides the candidate drugs against drug-tolerant persister cells in colorectal cancer

Yosui NOJIMA *¹, Ryoji YAO², Takashi SUZUKI¹

¹Center for Mathematical Modeling and Data Science, Osaka University
²Department of Cell Biology, Japanese Foundation for Cancer Research
( * E-mail: nojima@sigmath.es.osaka-u.ac.jp )

The inactivation of the APC gene is a crucial early event in colorectal cancer (CRC) development. Familial adenomatous polyposis (FAP) is a hereditary syndrome characterized by numerous adenomas in the colon, which significantly increase CRC risk due to an autosomal dominant mutation in the APC gene. The carcinogenesis mechanism in FAP mirrors that of sporadic CRC.

Previously, we evaluated the efficacy of anticancer drugs using organoids derived from benign and malignant tumors in FAP patients. The results showed that organoids from malignant tumors were resistant to the MEK inhibitor trametinib, likely due to the presence of drug-tolerant persister (DTP) cells in the cancer tissues.

Single-cell RNA sequencing (scRNA-Seq) is a powerful technology that allows for high-resolution analysis of gene expression in individual cells, offering new insights into cancer biology. Machine learning (ML), a branch of artificial intelligence, leverages statistical techniques to learn from data and is increasingly used in cancer diagnosis, prognosis, and treatment.

In this study, we conducted scRNA-Seq on FAP-derived organoids and built ML models using public data to identify DTP cells resistant to MEK inhibitors. Additionally, we identified candidate drugs against DTP cells in FAP organoids using public drug sensitivity data and validated the effects of these drugs using a cell viability assay.

This is the first study to demonstrate that ML models can identify DTP cells and propose a novel strategy for identifying candidate drugs against DTP cells.