Software Benchmark—Classification Tree Algorithms for Cell Atlases Annotation Using Single-Cell RNA-Sequencing Data

Classification tree is a widely used machine learning method. It has multiple implementations as R packages; rpart, ctree, evtree, and C5.0. The details of these are not the same, hence their performances differ from one application to another. We interested in performance classification cells using single-cell RNA-Sequencing data. In this paper, we conducted benchmark study 22 Single-Cell RNA-sequencing data sets. Using cross-validation, compare packages’ prediction based on Precision, Recall, F1-score, Area Under Curve (AUC). also compared Complexity Run-time packages. Our shows that rpart evtree have best Precision; F1-score AUC; C5.0 prefers more complex trees; consistently much faster than others, although its complexity often higher others.