Fig. 3: EvoWeaver is sufficiently accurate to hierarchically classify functional associations.
From: EvoWeaver: large-scale prediction of gene functional associations from coevolutionary signals
a The confusion matrix of five-level classifications indicates that EvoWeaver’s ensemble predictions (i.e., Random Forest) rarely confuse proteins within the same module with those from different modules. Values represent the percent of each actual class classified to each predicted class. b The best performing algorithm from each category on the Modules benchmark was also assigned greater feature importance by the random forest model in hierarchical classification. All features were important in the ensemble’s predictions, further underscoring the benefit of using multiple coevolutionary signals. Overlaid points denote importance from each of the five train/test folds. c A group of proteins randomly selected from hierarchical clustering exactly matches an existing tightly linked set of modules from KEGG. d EvoWeaver’s ensemble predictions for genes involved in prodigiosin biosynthesis generally match experimentally verified connections in KEGG. Note that pigA, pigJ, pigH, pigM, and pigF belong to both modules. Source data are provided as a Source Data file. Created in BioRender. Lakshman, A. (2025) https://BioRender.com/k26g262.
