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

BMC Bioinformatics

Table 3 PPIN triangle success in MIPS complex prediction.

CD = structural SDDI, protein-SCOP domain assignments > confidence threshold
Confid.thres.	Gavin06MATRIX	Gavin06SPOKE	Krogan06
0	254/2832	143/521	254/1182
40	160/2053	91/367	168/939
50	70/1192	42/215	99/679
60	44/786	29/152	60/467
70	38/704	23/146	55/418
80	36/639	21/130	40/337
90	35/601	21/124	39/306
CD = threading, protein-SCOP domain assignments > confidence threshold
Confid.thres.	Gavin06MATRIX	Gavin06SPOKE	Krogan06
medium	0/24	0/1	1/12
high	56/296	30/69	68/112
certain	205/4290	123/416	219/1099

Fractions denote True Positive PPIs /All triangle PPIs. In triangles where the complementary data (CD) are SDDI structural information or threading, SCOP domain families first need to be assigned to proteins based on confidence. The confidences for protein-SCOP domain assignments are derived, for structural SDDIs based on BLAST sequence alignment similarity, and for threading they are provided by the GTD threading database. Considering different confidence thresholds for protein-SCOP domain assignments affects the MIPS complex prediction success rate.

ISSN: 1471-2105