Identification of donor splice sites using support vector machine: a computational approach based on positional, compositional and dependency features

Table 3 Performance accuracies of different methods in predicting donor splice sites using NN269 dataset

Approaches	AUC-ROC	AUC-PR	Type of kernel used
MM1-SVM	97.62	89.58	Polynomial
LIK-SVM	98.04	92.65	Locally improved kernel
WD-SVM	98.50	92.86	Weighted degree kernel
WDS-SVM	98.13	92.47	Weighted degree shift kernel
EFFECT	98.20	92.81	–
Proposed	96.53	93.54	Radial basis function

It can be seen that WD-SVM achieved higher value of AUC-ROC as compared to the others, whereas the AUC-PR is highest for the proposed approach. MM1-SVM achieved lowest accuracies both in terms of AUC-ROC and AUC-PR

ISSN: 1748-7188