Fig. 6: EvoWeaver partly identifies biochemical pathway connectivity.
From: EvoWeaver: large-scale prediction of gene functional associations from coevolutionary signals
EvoWeaver’s pairwise scores from component algorithms provide a ranking of functional association drawn from alternative categories of coevolutionary signal (colors). EvoWeaver combines 12 component scores into a single ensemble prediction for each pair of gene groups. The strongest predicted connection for every gene group shows high consistency (solid lines) with the actual connectivity of KEGG pathways (arrows). Discrepancies (dashed lines) between predicted and actual connections are often caused by EvoWeaver incorrectly linking consecutive gene groups, such as AcbO-AcbL-AcbN in Acarbose Biosynthesis or ArgB-ArgC-ArgD in Ornithine Biosynthesis. Component predictors are connected according to the gene group with the highest mean rank among all algorithms in a category. Ensemble predictions are determined by connecting each gene group to the gene group with the highest probability of “Direct Connection” in the Multiclass Random Forest model.
