. 2016 Oct 22;33(3):347–353. doi: 10.1093/bioinformatics/btw656

Table 3.

Single- and multi-label performance with different combinations of input features on the SCEXP2016 dataset by adopting a 10-fold cross-validation procedure

	Multi-label prediction					Single-label prediction
Input features	ACC^ml	mlACC	mlPRE	mlREC	mlF1	ACC^sl(I)	ACC^sl(O)	ACC^sl(S)	ACC^sl(L)	ACC^sl(M)	ACC^sl(P)	ACC^sl
Basic	0.48	0.56	0.56	0.61	0.59	0.33	0.26	0.60	0.61	0.57	0.48	0.58
Basic+target (predicted)	0.62	0.75	0.76	0.90	0.83	0.62	0.57	0.89	0.89	0.82	0.80	0.89
Basic+mem (predicted)	0.60	0.74	0.75	0.89	0.82	0.67	0.65	0.84	0.85	0.83	0.75	0.85
Basic+target+mem (predicted)	0.63	0.77	0.79	0.93	0.85	0.66	0.63	0.90	0.89	0.84	0.81	0.91
Basic+target+mem (observed)	0.73	0.82	0.84	0.93	0.89	0.81	0.75	0.97	0.97	0.92	0.86	0.96

Basic = PSSM + Hydrophobicty; target = [p(c),p(t)]; mem = [p(s),p(m),p(r)]. Scoring indexes are defined as in Section 2.3. In single-label scoring indexes, I, O, S, L, M and P stand for inner membrane, outer membrane, stroma, thylakoid lumen, thylakoid membrane and plastoglobule, respectively.