Making the most of high-throughput protein-interaction data

Table 1 Overview of seven Y2H and five AP-MS experiments

Reference	VB	CB	TB	VP	VBP	VBP/BP	TI	TI/VB	REC	UNR
Ito et al. [1]	1,522		6,604	2,493	773	0.51	4,524	3.0	75	803
Cagney et al. [2]	19		31	40	11	0.58	54	2.9	3	4
Tong et al. [3]	20		22	59	5	0.25	115	5.8	1	1
Hazbun et al. [4]	66		100	1,940	28	0.42	2,524	38	4	13
Zhao et al. [5]	1		1	90	0	0.00	90	90	0	0
Uetz et al. Experiment 1 [6]	508		6,604	630	142	0.28	952	1.9	10	47
Uetz et al. Experiment 2 [6]	139		192	400	36	0.26	524	3.8	18	7
Gavin et al. [7]	455	600	725	1,179	271	0.60	3,419	7.5	192	314
Ho et al. [8]	493	589	1,739	1,316	231	0.47	3,687	7.5	69	297
Krogan et al. [9]	153	165	165	483	151	0.99	1,132	7.4	89	157
Gavin et al. [10]	1,752	1,993	6,466	1,790	991	0.57	19,105	10.9	1,077	4,297
Krogan et al. [11]	2,264	2,357	4,562	5,323	2,226	0.98	63,360	28.0	1,969	34,363

VB, the number of viable baits; CB, the number of cloned (hybridized) baits, if available; TB, the total number of baits that the experimenters were initially aiming at; VP, the number of viable prey; VBP, the number of proteins observed as both bait and prey; TI, the total number of interactions observed; REC, the number of reciprocated interactions between proteins that were observed as both bait and prey; UNR, the number of unreciprocated interactions between proteins that were observed as both bait and prey. Not all of the experiments were genome-wide - some were focused on particular aspects of the cellular machinery [2-5,9]. Even in the so-called genome-wide studies [1,6-8,10,11], however, the viable baits cover only around a third of the yeast genes. This means that the largest part of interaction space by far, containing interactions between proteins not used as baits, was not sampled in any of these experiments. We can also see that TI/VB, the average number of interactions per viable bait, varies markedly between experiments. In the more focused studies, this will certainly be a result of different criteria for the selection of baits. In the genome-wide screens it may indicate the application of different, experiment-specific cutoffs.

ISSN: 1474-760X