Fig. 2From: Evaluation of some aspects in supervised cell type identification for single-cell RNA-seq: classifier, feature selection, and reference constructionPrediction accuracy gains/losses with different reference sizes. A 0–1000 cells. B 1000–5000 cells. C 5000–10,000 cells. D 10,000+ cells. scmap ranks 1st in A and 2nd in B, but when cell number increases, scmap ranks 6th both in C and D. MLP ranks steadily as 1st or 2nd. SVM with RBF kernel first ranks 6th in A but later ranks 1st in D. The F-test on reference data ranks steadily as 1st among all feature selection methodsBack to article page