Test partitioning methods


In the process of recursively splitting the similarity matrix in binary cut algorithm, in each iteration step, the current matrix is partitioned into two groups using PAM as default. Here we compare following partitioning methds: k-means++, PAM and hierarchical clustering with 'ward.D2' method, on 500 random GO lists.

Figure 1.Compare clustering results. Left panel: The difference score, number of clusters and the block mean of different clusterings. Right panel: Concordance between clustering methods. The concordance measures how similar two clusterings are. The definition of the concordance score can be found here.

Table 1.Number of clusters identified by each clustering method. Numbers in the table indicate the number of clusters. The numbers inside the parentheses are the number of clusters with size >= 5.

run partition_by_pam partition_by_kmeanspp partition_by_hclust Details
1 23 (11) 17 (10) 18 (10) view
2 21 (11) 16 (9) 16 (11) view
3 17 (10) 18 (8) 17 (10) view
4 21 (12) 24 (10) 17 (13) view
5 1 (1) 21 (10) 17 (9) view
6 1 (1) 13 (6) 16 (10) view
7 19 (10) 21 (6) 16 (9) view
8 18 (11) 21 (11) 10 (10) view
9 23 (11) 19 (8) 18 (10) view
10 6 (6) 18 (11) 20 (11) view
11 21 (10) 19 (11) 14 (7) view
12 1 (1) 19 (8) 18 (9) view
13 17 (11) 12 (11) 21 (10) view
14 23 (11) 21 (8) 23 (10) view
15 22 (10) 19 (9) 18 (10) view
16 1 (1) 19 (10) 19 (11) view
17 17 (9) 12 (8) 12 (8) view
18 21 (10) 19 (8) 18 (10) view
19 26 (12) 16 (8) 18 (6) view
20 22 (10) 17 (10) 16 (11) view
21 25 (10) 18 (9) 16 (10) view
22 20 (8) 18 (10) 18 (9) view
23 18 (9) 17 (9) 19 (9) view
24 25 (12) 19 (11) 18 (9) view
25 19 (9) 19 (8) 15 (9) view
26 20 (12) 19 (11) 17 (10) view
27 16 (8) 16 (9) 16 (9) view
28 18 (11) 15 (8) 16 (9) view
29 18 (7) 19 (7) 17 (8) view
30 24 (7) 19 (9) 16 (8) view
31 25 (14) 30 (11) 22 (12) view
32 21 (11) 20 (11) 20 (11) view
33 15 (10) 17 (9) 12 (12) view
34 26 (10) 15 (12) 17 (8) view
35 28 (10) 24 (9) 28 (10) view
36 29 (11) 20 (8) 21 (12) view
37 24 (11) 1 (1) 22 (11) view
38 19 (10) 20 (10) 20 (10) view
39 6 (6) 20 (9) 16 (9) view
40 22 (10) 18 (9) 19 (10) view
41 20 (9) 22 (8) 19 (9) view
42 8 (8) 20 (11) 17 (8) view
43 5 (5) 17 (8) 17 (9) view
44 23 (8) 18 (10) 19 (8) view
45 18 (9) 14 (9) 13 (10) view
46 22 (11) 17 (9) 15 (9) view
47 20 (11) 17 (9) 15 (10) view
48 16 (10) 16 (8) 15 (8) view
49 16 (6) 15 (9) 16 (7) view
50 26 (9) 17 (8) 16 (7) view
51 3 (3) 17 (9) 14 (8) view
52 5 (5) 19 (9) 18 (9) view
53 5 (5) 19 (9) 17 (9) view
54 20 (10) 17 (10) 17 (12) view
55 16 (12) 16 (12) 13 (11) view
56 20 (9) 14 (9) 21 (8) view
57 22 (12) 18 (9) 16 (10) view
58 14 (7) 19 (9) 17 (7) view
59 12 (7) 3 (3) 11 (8) view
60 21 (11) 19 (9) 17 (11) view
61 25 (8) 22 (7) 17 (9) view
62 25 (10) 17 (8) 16 (11) view
63 19 (10) 21 (9) 18 (11) view
64 16 (9) 17 (9) 16 (9) view
65 5 (5) 19 (10) 15 (10) view
66 5 (5) 16 (11) 16 (9) view
67 17 (10) 16 (10) 15 (10) view
68 27 (9) 21 (11) 17 (8) view
69 21 (11) 18 (9) 20 (9) view
70 23 (11) 21 (8) 18 (8) view
71 5 (5) 14 (9) 12 (10) view
72 23 (11) 19 (9) 16 (9) view
73 21 (12) 9 (9) 17 (10) view
74 8 (8) 16 (9) 11 (11) view
75 19 (13) 14 (13) 16 (11) view
76 15 (9) 12 (8) 11 (11) view
77 23 (11) 18 (10) 18 (10) view
78 19 (9) 18 (7) 21 (11) view
79 24 (8) 16 (9) 21 (9) view
80 27 (11) 18 (11) 18 (10) view
81 24 (12) 21 (10) 20 (12) view
82 5 (5) 21 (8) 25 (12) view
83 11 (5) 14 (10) 21 (10) view
84 14 (9) 17 (7) 15 (7) view
85 20 (6) 24 (6) 22 (8) view
86 20 (10) 20 (9) 18 (9) view
87 27 (8) 21 (7) 18 (8) view
88 17 (14) 20 (10) 25 (9) view
89 5 (5) 14 (6) 17 (9) view
90 22 (11) 20 (8) 20 (11) view
91 16 (10) 16 (8) 17 (8) view
92 1 (1) 20 (7) 19 (11) view
93 23 (13) 16 (9) 18 (10) view
94 21 (10) 23 (10) 13 (12) view
95 21 (10) 16 (7) 22 (11) view
96 18 (11) 19 (11) 16 (9) view
97 20 (9) 17 (9) 16 (7) view
98 25 (11) 18 (8) 20 (11) view
99 20 (11) 18 (10) 17 (10) view
100 16 (10) 11 (9) 16 (9) view