Table 1.

Prediction performance comparisons between PredCRP-model and the SVM-based methods with various feature sets on the CRPS-TRN dataset.

Feature set	No. of features	SPE	SEN	MCC	ACC
PredCRP-model	12	1.00	0.92	0.95	0.98
Informative 4-mer motifs	8	0.99	0.38	0.52	0.86
Informative location features	4	1.00	0.29	0.49	0.85
All features (baseline)	380	0.98	0.29	0.41	0.83
Composition descriptor	320	0.98	0.17	0.26	0.81
Location-dependent descriptor	17	0.99	0.29	0.45	0.84
Location-independent descriptor	363	0.99	0.17	0.31	0.81