File:Fpls-02-00043-g004.png

English: Matthew’s correlation coefficient plots presenting the performance of the predictors constructed for 10 subcellular compartments. The investigated compartments were: the vacuole, the peroxisome, the cell wall, the cytosol, the ER, the plastid, the Golgi apparatus, the nucleus, the plasma membrane, the mitochondrion. For each of the 10 compartments the prediction engines were built using: amino acid composition (AA), the top 20 expression features (T20 E), the top 20 mixed features selected from the amino acid composition and the expression data (T20 AA + E), the top 1,000 features selected from the expression features (T1000 E), the top 1,000 amino acid composition and expression features (T1000 AA + E) and the top 1,000 features selected from amino acid composition, dipeptide composition and expression data (T1000 AA + D + E). For each predictor the Matthews’ correlation coefficients from the 5 cross-validation loops are visualized as a box plot.