PrePPI: A Structure Informed Proteome-wide Database of Protein-Protein Interactions

Abstract

We present an updated version of the Predicting Protein-Protein Interactions (PrePPI) webserver which predicts PPIs on a proteome-wide scale. PrePPI combines structural and non-structural evidence within a Bayesian framework to compute a likelihood ratio (LR) for essentially every possible pair of proteins in a proteome; the current database is for the human interactome. The structural modeling (SM) component is derived from template-based modeling and its application on a proteome-wide scale is enabled by a unique scoring function used to evaluate a putative complex. The updated version of PrePPI leverages AlphaFold structures that are parsed into individual domains. As has been demonstrated in earlier applications, PrePPI performs extremely well as measured by receiver operating characteristic curves derived from testing on E. coli and human protein-protein interaction (PPI) databases. A PrePPI database of ∼1.3 million human PPIs can be queried with a webserver application that comprises multiple functionalities for examining query proteins, template complexes, 3D models for predicted complexes, and related features (https://honiglab.c2b2.columbia.edu/PrePPI). PrePPI is a state-of-the-art resource that offers an unprecedented structure-informed view of the human interactome.

Keywords: alphafold models; database; non-structural evidence; protein-protein interactions; structural modeling.

Figure 1:

PrePPI’s structural modeling (SM) pipeline: Structures for query proteins, QA and QB, are taken from the AlphaFold Protein Structure Database [13] and parsed into domains with definitions from the Conserved Domain Database (CDD) [22]. Structural neighbors in the PDB [3] for full length protein and domain structures are obtained from the ska structural alignment program [31]. If structural neighbors of two query proteins appear together in a PDB complex, this structure defines a template, NA1:NB3, used to create a structure-based sequence alignment with which an interface for the query proteins, QA:QB, is evaluated based on the overlap of the query and template residues [1]. The interaction is then scored based on a number of features [1, 2] and trained on the HINT HQ-LC database [10], as the positive set, and a negative set described in Methods to produce a fully connected Bayesian network used to evaluate the model.