PEPPI (Pipeline for the Extraction of Predicted Protein-protein
Interactions) is a computational program for protein-protein interaction (PPI)
prediction. Given a pair of protein amino acid sequences, PEPPI predicts the likelihood
of direct, physical interaction for those sequences through several independent prediction
methods, including protein structual homology by multimeric threading, protein sequence homology
by BLAST search through high-throughput experimental data, functional association from the
STRING database, and machine learning-based classification. Scores from each of these
approaches are combined through a naive Bayesian consensus model into a final likelihood ratio
expressing the probability of interaction relative to the probability of non-interaction.
Note: PEPPI is efficiently implemented for the modeling of proteome-wide interaction networks;
for this purpose we recommend users utilize the standalone package below. This webserver limits
users to one pair of proteins interacting at a time.