Fig. 3.

Predictive performance of NRTPredictor. A ROC curves for NRTPredictor on training set. B The confusion matrix shows the accuracy of NRTPredictor using 110 genes from MIC_SVM algorithm on test dataset. C and D The clustering effect on 3463 cells was evaluated using 110 marker genes and all genes (C represents all genes, D represents the 110 marker genes). Each point represents a sample in the dataset, and different categories of samples are given different colors. E Correlation analysis of six rice root cell subpopulations at the level of 110 marker genes. F Boxplots of Mean expression levels (Normalization of raw read count) of 110 marker genes in six rice root cell subpopulations