Table 2.

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

Feature set	No. of features	SPE	SEN	MCC	ACC
PredCRP-model	12	0.95	0.83	0.79	0.93
Informative 4-mer motifs	8	0.97	0.25	0.36	0.82
Informative location features	4	0.98	0.25	0.36	0.82
All features (baseline)	380	0.98	0.17	0.26	0.80
Composition descriptor	320	0.93	0.33	0.33	0.80
Location-dependent descriptor	17	1.00	0.08	0.26	0.80
Location-independent descriptor	363	0.93	0.33	0.33	0.80