Single cell computational tools comparison to detect epithelial tumoral cells

Abstract

Single-cell RNA sequencing offers a unique view of the transcriptome. With a lot of different applications on this project, we have the objective of identifying the different cell types and tumoral cells on all the datasets with the Python tool to process single-cell RNA sequencing data, Scanpy. The VHIO’s bioinformatics team has performed the same analysis, but with Seurat, the equivalent tool from R, another programming language. Another objective of this project is to compare the two more commonly used packages to analyze this data with both pipelines in order to find the most efficient way to work with single-cell data. To perform the analysis, we used public data from Gene Expression Omnibus pancreas ductal adenocarcinoma. The development of the Scanpy pipeline was mostly following Scanpy’s tutorial, and the whole project was performed on the VHIO’s cluster. The results obtained with the annotation of the cell type were correlated with the marker genes characteristic of each cell type, resulting in a satisfactory result. The comparison was done by segmenting the process on different scripts to compare the various stages between Python and R and registering the memory RAM used and the elapsed time. The result was R using fewer resources than Python, and the integration of all the samples exacerbated the gap between performances.