Fig. 3From: A novel codon-based de Bruijn graph algorithm for gene construction from unassembled transcriptomesPerformance evaluation of SVM filtration. a Performance of SVM on filtering false-positive ORFs. The subgraph number and the ratio of false-positive nodes in each subgraph were calculated before and after SVM filtration. b Comparison between the two codon-based de Bruijn graphs before and after SVM filtration. c Sensitivity of true ORF recovery after SVM filtration using four simulated datasets with different read lengths (100, 300, 500, and 800Â bp). d Classification results of SVM filtration through reducing the high-dimensional feature vectors by principal component analysis. Red dots indicate true ORFs belonging to CDSs, whereas black dots represent false ORFs derived from non-CDSsBack to article page