Triangle network motifs predict complexes by complementing high-error interactomes with structural information

BMC Bioinformatics

Table 4 Individual ability of various datatypes to predict MIPS complexes.

Network	Number of nodes ^a	Number of edges ^b	Node overlap with MIPS ^c	MIPS nodes not in network ^d	Network nodes not in MIPS ^e	Nodes in edge overlap ^f	Edges in edge overlap ^g	MIPS nodes not in edge overlap ^h	MIPS edges not in network ⁱ	Network edges not in MIPS ^j
Gavin06MATRIX	2551	93881	584	791	1967	554	305	821	247	3989
Gavin06SPOKE	2551	22452	584	791	1967	535	232	840	320	950
Krogan06	3670	14291	1031	344	2639	994	359	381	928	2225
SDDI ^k	1551	42222	515	860	1036	500	182	875	207	7375
Threading	4019	95935	1037	338	2982	1017	219	358	1129	11938
Literature co-occurrence ^l	96379	170638	1056	319	95323	1235	491	140	1299	2129
Domain co-occurrence ^m	3560	158704	1042	333	2518	1038	287	337	940	6064
All above combined	100443	504242	1351	24	99092	1358	979	17	1048	24156

Symbols below denote N, nodes; E, edges; |·|, set cardinality; ∩, intersection; -, set difference; ×, cross product
a |N_network|
b |E_network|
c |N_MIPS∩ N_network|
d |N_MIPS- N_network|
e |N_network- N_MIPS|
f |Nodes in E_MIPS∩ E_network|
g |E_MIPS∩ E_network|
h |N_MIPS- Nodes in E_MIPS∩ E_network|
i |(E_MIPS∩ (N_network× N_network)) - E_network|
j |(E_network∩ (N_MIPS× N_MIPS)) - E_MIPS|
k Structural domain-domain interactions
l PubMed literature co-occurrences of protein mentions
m Pfam domain co-occurrences in IntAct PPIs
This table shows the MIPS overlaps with other network datasets (shown in the first column), indicating the ability of the various networks to predict MIPS. The MIPS network has number of protein nodes |N_MIPS| = 1375 and number of edges |E_MIPS| = 2099.

ISSN: 1471-2105