. 2015 Aug 27;16:271. doi: 10.1186/s12859-015-0713-y

Table 2.

Results of predicting the missing BP functions of partially annotated Yeast proteins (N=5914, $| T | = 2979$ )

Metric	m	dRW-kNN	dRW	ITSS	PILL	Naive
MacroF1	1	93.14 ± 0.13	93.61 ± 0.05	91.66 ± 0.09	91.52 ± 0.15	1.99 ± 0.00
	3	82.72 ± 0.25	83.29 ± 0.13	80.14 ± 0.14	79.77 ± 0.16	2.01 ± 0.00
	5	74.67 ± 0.22	75.92 ± 0.17	71.16 ± 0.33	70.96 ± 0.22	2.03 ± 0.00
AvgROC	1	99.88 ± 0.01	99.93 ± 0.00	98.24 ± 0.02	98.77 ± 0.03	45.88 ± 0.00
	3	99.55 ± 0.02	99.59 ± 0.02	94.44 ± 0.08	96.36 ± 0.15	45.88 ± 0.00
	5	99.01 ± 0.02	98.89 ± 0.03	90.48 ± 0.17	93.83 ± 0.06	45.88 ± 0.00
1-RankLoss	1	99.96 ± 0.00	99.97 ± 0.00	98.99 ± 0.02	99.81 ± 0.01	91.13 ± 0.00
	3	99.47 ± 0.03	99.17 ± 0.03	96.89 ± 0.05	99.23 ± 0.03	91.04 ± 0.00
	5	98.20 ± 0.02	97.63 ± 0.03	93.99 ± 0.10	98.42 ± 0.05	90.95 ± 0.01
Fmax	1	97.97 ± 0.00	98.08 ± 0.00	97.90 ± 0.00	97.91 ± 0.00	36.96 ± 0.00
	3	93.99 ± 0.02	93.92 ± 0.01	93.66 ± 0.02	93.61 ± 0.00	36.86 ± 0.00
	5	90.25 ± 0.03	89.88 ± 0.00	89.66 ± 0.02	89.41 ± 0.00	36.84 ± 0.03
RAccuracy	1	38.75 ± 0.66	44.68 ± 0.21	12.41 ± 0.48	21.65 ± 0.37	37.51 ± 0.94
	3	39.75 ± 0.13	36.08 ± 0.24	23.02 ± 0.06	22.27 ± 0.37	37.84 ± 0.75
	5	40.13 ± 0.43	33.58 ± 0.24	27.39 ± 0.29	23.92 ± 0.08	37.69 ± 0.37
Coverage ↓	1	78.24 ± 0.95	66.34 ± 1.10	405.01 ± 9.56	232.52 ± 4.58	1585.06 ± 0.99
	3	191.34 ± 4.88	234.61 ± 4.92	943.54 ± 10.84	524.50 ± 12.33	1605.22 ± 0.95
	5	340.09 ± 4.55	469.13 ± 9.17	1412.81 ± 9.85	806.23 ± 18.18	1625.35 ± 3.17

The numbers in boldface denote the best (or comparable best) statistically significant performance (according to a t-test at 95 % significance level). ↓ means the lower the value, the better the performance. m is the number of missing functions for a protein, N _m is the total number of missing functions, and $| T_{m}^{0} |$ is the number of the second kind of missing functions of N proteins for a given m. m=1, $| T_{1}^{0} | = 25$ , N ₁=4705; m=3, $| T_{3}^{0} | = 106$ , N ₃=14079; m=5, $| T_{5}^{0} | = 209$ , N ₅=23299