A domain-based approach to predict protein-protein interactions

BMC Bioinformatics

Table 2 Explanation ratios of the MIPS yeast datasets

No of Parameters	Training set	PPI	Explanation Ratio (%)^a	Accuracy	Precision
867	Inclusive	Positive	95.6	96.4	38.9
		Negative	96.1
344	Closed	Positive	99.3	98.7	90.6
		Negative	98.7
2466	Inclusive	Positive	96.2	95.7	24.3
		Negative	95.7
1216	Closed	Positive	99.0	96.6	60.8
		Negative	96.4
5095	Inclusive	Positive	97.3	95.6	19.3
		Negative	95.6
3060	Closed	Positive	99.3	95.9	56.5
		Negative	95.6
Deng et al. with	Inclusive	Positive	95–98	93–95	24–30
867 pmts ^b		Negative	93–95
	Closed	Positive	91–93	89–90	54–55
		Negative	89–90
Random with	Inclusive	Positive	61.0	36.0	2.2
867 pmts		Negative	35.4

^a Explanation ratio is the ratio of successful predictions to the overall number of entries in a particular list. We note that the explanation ratios for the positive and negative PPI lists respectively correspond to the sensitivity and specificity with Lin et al. definitions.
^b Calculated performance metrics depend on the false positive and negative prediction rates used in the MLE algorithm as well as on the cutoff for positive and negative PPI assignments. Therefore, we report the range of obtained percentages that were obtained when various prediction rates were used in the MLE algorithm. Whole set of results can be found in Supplementary Table 1 [see Additional file].

ISSN: 1471-2105