Skip to main content
. 2021 Nov 9;12:6454. doi: 10.1038/s41467-021-26792-w

Fig. 1. Method comparison for prediction of functional interactions.

Fig. 1

Our model (MLPP) was compared against other phylogenetic profiling approaches in terms of the receiver operator curve (ROC) and area under the curve (AUC) in predicting pairs of functionally interacting genes. Additionally, MLPP could predict the interaction context - complex co-occurrence, or one of 12 top-level pathways from Reactome. The model outperformed other approaches in predicting functional interaction (A, B) and the interaction context (B) when compared in 5-times cross-validation. Error bars denote the 95% confidence intervals using 1000 bootstrap samples. MLPP—machine-learning phylogenetic profiling, NPP—normalized phylogenetic profiling, SVD-Phy—singular value decomposition phylogenetic profiling, PPP—PrePhyloPro, Hamming—binarized phylogenetic profiling with Hamming distance. Source data are provided as a Source Data file.