Increasing the number of datasets for MEM queries improves prediction of Mini Chromosome Maintenance (MCM) subunits. As additional datasets are incorporated for MEM analysis, MCM complex subunits show more consistent expression patterns as measured by median distance between subunits in MEM ranked lists of most correlated genes (decreasing bar height). According to one-sided Kolmogorov-Smirnov tests, MEM analysis with different numbers of datasets (left bars) significantly outperforms correlation (rightmost bar). In addition, MEM analysis for all the 145 selected datasets gives improved results compared to plain correlation across the concatenated dataset (light blue and orange lines).
Adler et al. Genome Biology 2009 10:R139 doi:10.1186/gb-2009-10-12-r139