Skip to main content

Table 2 Summary of the results for mislabelled negative sites

From: Positive-unlabelled learning of glycosylation sites in the human proteome

Year

 

C-linked

N-linked

O-linked

2010

N1a

0

237 (26.04%)

22 (91.67%)

P1

11.76%

3.38%

1.26%

P2

14.86%

3.41%

1.28%

P3

12.07%

3.39%

1.28%

2013

N1

0

119 (36.73%)

32 (19.82%)

P1

8.35%

3.01%

1.22%

P2

9.36%

4.36%

1.24%

P3

8.62%

3.97%

1.22%

2016

N1

0

99 (34.62%)

32 (19.82%)

P1

6.51%

3.09%

1.11%

P2

7.15%

4.68%

1.13%

P3

6.63%

3.83%

1.13%

  1. Note: a) N1, numbers and percentages of mislabelled non-glycosylation sites and their percentages as compared with previous collection years; b) P1, the actual class probability of glycosylation sites; c) P2: the prior probability of glycosylation sites estimated by the Elkan-Noto algorithm; d) P3: the prior probability of glycosylation sites estimated by the AlphaMax algorithm