Method
|
Enriched Clusters
|
# NEC
|
% NEC
|
# NEC S
|
% NEC S
|
---|
PPI
|
35/64 (54.69%)
|
3274.0
|
53.69%
|
1703.0
|
27.93%
|
3.5
|
56/91 (61.54%)
|
570.0
|
9.35%
|
468.0
|
7.68%
|
4.0
|
97/142 (68.31%)
|
1155.0
|
18.95%
|
915.0
|
15.01%
|
4.5
|
144/215 (66.98%)
|
1869.0
|
30.66%
|
1415.0
|
23.21%
|
5.0
|
96/174 (55.17%)
|
2785.0
|
45.69%
|
1724.0
|
28.28%
|
5.5
|
56/93 (60.22%)
|
4067.0
|
66.72%
|
1783.0
|
29.25%
|
6.0
|
51/81 (62.96%)
|
4155.0
|
68.16%
|
1667.0
|
27.35%
|
PPI
|
39/69 (56.52%)
|
3367.0
|
55.21%
|
1782.0
|
29.22%
|
3.5
|
55/91 (60.44%)
|
495.0
|
8.12%
|
463.0
|
7.60%
|
4.0
|
97/142 (68.31%)
|
1155.0
|
18.95%
|
915.0
|
15.01%
|
4.5
|
144/215 (66.98%)
|
1869.0
|
30.66%
|
1415.0
|
23.21%
|
5.0
|
95/174 (54.60%)
|
2686.0
|
44.06%
|
1676.0
|
27.49%
|
5.5
|
60/106 (56.60%)
|
3978.0
|
65.26%
|
1862.0
|
30.54%
|
6.0
|
66/96 (68.75%)
|
4077.0
|
66.88%
|
1680.0
|
27.56%
|
- The numbers above the double line are for cutting the Walktrap dendrogram at 200 clusters; the numbers below the double line are for cutting the Walktrap dendrogram at 300 clusters. NEC= “Nodes in Enriched Clusters”. We calculate %NEC in two settings: %NEC is enrichment in the GO hierarchy with terms above the fifth level filtered out, and %NEC S uses the same filtered GO hierarchy, but then only gives a node credit if there is a match between one of the node’s labels and one of the terms for which there is GO enrichment for the cluster. In both cases, for the S statistic the best DSD threshold is 5.5, at which performance is slightly better than running Walktrap directly on the PPI network. For cutoffs of both 200 and 300 nodes, DSD+Walktrap is slightly better than Walktrap in the NEC measure, and in both cases the DSD version produces slightly more and smaller clusters
- Bolded values represent the best values achieved for the %NEC and %NEC S statistics comparing the PPI network and different DSD detangling thresholds