Evaluation of Single-Cell RNA-Seq Clustering Algorithms on Cancer Tumor Datasets: A Summary

Mahalanabis, Alaina, Andrei L. Turinsky, Mia Husić, Erik Christensen, Ping Luo, Alaine Naidas, Michael Brudno, Trevor Pugh, Arun K. Ramani, and Parisa Shooshtari.

Cancer is a complex disease involving a mix of different cell types that need to be understood to improve diagnosis and treatment. Recent advancements in single-cell RNA sequencing (scRNA-seq) allows researchers to study the cell types that compose various tumours and identify key similarities or trends. In analyzing scRNA-seq datasets, the cell types that compose tumours are grouped by specialized programs, a process known as clustering. However, selecting the best method for clustering these cells from scRNA-seq data is challenging, especially because different tools may be designed for specific cell types or purposes. 

To address this, we have developed a framework to evaluate 15 different scRNA-seq clustering programs, testing them on eight diverse cancer datasets, including brain, breast, lung, colorectal, pancreatic cancers, leukemia, and melanoma. The results show that clustering cancer cells is particularly challenging due to the complexity of tumour environments. However, some methods perform better than others depending on the type of tumour microenvironment being analyzed. For example, the Seurat algorithm works well for non-cancerous cells, while Monocle and SC3 are more effective for cancerous cells.

By comparing clustering programs, the study helps researchers choose the best method for analyzing their cancer data, which could lead to better understanding of tumour biology and help in developing more personalized cancer treatments.

Link to full text: https://doi.org/10.1016/j.csbj.2022.10.029.