Figure 1

Balanced profiles outperforms the raw profiles. Compared to raw profiles with varying weights on the four phylogenetic scenarios of coevolution, balanced profiles with equal weight on those scenarios improve the prediction accuracy and identify many-fold more putative triplets as well. Triplets with GO terms refers to triplets where GO terms are available for all three genes, and putative triplets are those triplets where the three genes share common GO terms. The percentage of putative triplets in all triplets with GO terms is a measure of prediction accuracy. A: ΔU is a score to signal the significance of a triplet, ranging from -1 (very weak) to 1 (very strong). A bar at ΔU of 0.125 refers to the region [0.10, 0.15). The number of predicted triplets decreases exponentially with the increase of ΔU. And there is a strong positive correlation (R² = 0.99) between the % of putative triplets and ΔU, an evidence that ΔU indeed signals functional relatedness of genes. B: Ranked in the descending order of ΔU, the top triplets computed from balanced profiles consistently display a higher tendency to have functional relations within their three genes than the raw profile, indicating the benefit of giving equal weight to the four scenarios of coevolution between genes. In addition, substantially more triplets are produced from balanced profiles than the raw profiles without sacrifice on accuracy.