Model-based analysis of two-color arrays (MA2C)

Table 1 Comparison of MA2C with other algorithms using a spike-in experiment with a total of 96 regions and 47 unique non-overlapping regions

Algorithm	CHIP_ID	PPV	Sensitivity	Unique	Correlation
ChIPOTle	49875	71%	85%	40	0.72
	49880	69%	98%	47	0.76
	49883	73%	98%	47	0.79
MPeak	49875	100%	91%	46	0.74
	49880	96%	89%	46	0.71
	49883	98%	89%	46	0.79
MA2C	49875	99%	91%	47	0.78
(C = 2 normalized)	49880	96%	94%	47	0.79
	49883	99%	95%	47	0.81
	All 3	96%	96%	47	0.81
MA2C	49875	99%	92%	46	0.77
(Global median-scaled)	49880	100%	93%	46	0.79
	49883	99%	92%	46	0.81
	All 3	100%	95%	47	0.80

PPV (positive predictive value) = no. of true positive peaks/no. of total peaks. Sensitivity = no. of detected true positive regions/96. Unique = number of unique regions found. Correlation = correlation coefficient of the spike-in log fold-changes and algorithm-assigned scores for the 47 unique regions.

ISSN: 1474-760X