Figure 1.
The effect of parameter settings on the prediction performance of Presep for Type I (A) and Type II (B) PseAAC modes.
The x-axis represents the weight factor (w), and the y-axis represents the lambda parameter (λ). Colour changes indicate differences in the Matthews correlation coefficient (MCC).
Table 1.
Prediction performance of Presep with different parameters.
Figure 2.
The average prediction accuracy calculated cumulatively with RI above a given value.
This result was obtained using RF with a 20-fold cross validation.
Figure 3.
Prediction performance of Presep with different protein lengths.
The x axis represents the selected residues at N or C terminal that were used to predict the secretion propensity in the dataset. The y axis represents the prediction performance, which was evaluated by Matthews correlation coefficient (MCC). A Type I PseAAC coding scheme was used with a weight factor (w) of 0.05 and a lambda parameter (λ) of 19. Results are based on the Secreprot dataset with 20-fold cross validation.
Figure 4.
Correlation between the predicted secretion propensity, as determined by Presep, and the extracellular percentage (%) determined using the experimental method.
Table 2.
Predicted propensity and the experimental results on the six β-galactosidase.