A domain-based approach to predict protein-protein interactions

BMC Bioinformatics

Table 3 Cross verification with the yeast datasets

^a Explanation ratios were calculated by using the indicated yeast datasets as the testing data after the domain-domain interaction score parameters were optimized using the reported MIPS set as the training data in the DomainGA runs.
^b Explanation ratios were calculated by using the indicated closed Uetz et al. yeast datasets [1] as testing data. Uetz et al. datasets http://dip.doe-mbi.ucla.edu/dip/Download.cgi contain only positive PPIs so test statistics were computed only for the positive predictions. Yeast Core set originally contained 5952 positive PPIs of which 74 were retained after selecting the entries according to their relevance to the parameter set utilized in optimizations. In the Yeast Full dataset case the corresponding total and retained values were 17471 and 119.

ISSN: 1471-2105