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

Machine learning to predict de novo protein-protein interactions

Advances in machine learning for structural biology have dramatically enhanced our capacity to predict protein-protein interactions (PPIs). Here, we review recent developments in the computational prediction of PPIs, particularly focusing on innovations that enable interaction predictions that have no precedence in nature, termed de novo. We discuss novel machine learning algorithms for PPI prediction, including approaches based on co-folding and atomic graphs. We further highlight methods that learn from molecular surfaces, which can predict PPIs not found in nature including interactions induced by small molecules. Finally, we explore the emerging biotechnological applications enabled by these predictive capabilities, including the prediction of antibody-antigen complexes and molecular glue-induced PPIs, and discuss their potential to empower drug discovery and protein engineering.

Exploring the role of EBV in multiple sclerosis pathogenesis through EBV interactome

Background

Epstein-Barr virus (EBV) is a known risk factor for multiple sclerosis (MS), even though the underlying molecular mechanisms are unclear and engage multiple immune pathways. Furthermore, the ultimate role of EBV in MS pathogenesis is still elusive. In contrast, Cytomegalovirus (CMV) has been identified as a protective factor for MS.

Objectives

This study aims to identify MS-associated genes that overlap with EBV interactome and to examine their expression in immune and glial cell subtypes.

Methods

We used P-HIPSTer, GWAS, and the Human Protein Atlas (HPA) to derive data on the EBV interactome, MS-associated genes and single-cell gene expression in immune and glial cells. The geneOverlap and dplyr R packages identified overlapping genes. A similar analysis was done for CMV and Adenovirus as negative control. Metascape and GTEx analyzed biological pathways and brain-level gene expression; transcriptomic analysis was performed on glial cells and peripheral blood in MS and controls. All the analyses performed in this study were generated using publicly available data sets.

Results

We identified a "core" group of 21 genes shared across EBV interactome, MS genes, and immune and glial cells (p<0.001). Pathway analysis revealed expected associations, such as immune system activation, and unforeseen results, like the prolactin signaling pathway. BCL2 in astrocytes, MINK1 in microglia were significantly upregulated while AHI1 was downregulated in MS compared to controls.

Conclusions

Our findings offer novel insights into EBV and CMV interaction with immune and glial cells in MS, that may shed light on mechanisms involved in disease pathophysiology.

Deep learning methods for proteome-scale interaction prediction

Proteome-scale interaction prediction is essential for understanding protein functions and disease mechanisms. Traditional experimental methods are often limited by scale and complexity, driving the need for computational approaches. Deep learning has emerged as a powerful tool, enabling high-throughput, accurate predictions of protein interactions. This review highlights recent advances in deep learning methods for protein-protein and protein-ligand interaction screening, along with datasets used for model training. Despite the progress with deep learning, challenges such as data quality and validation biases remain. We also discuss the increasing importance of integrating structural information to enhance prediction accuracy and how structure-based deep learning approaches can help overcome current limitations, ultimately advancing biological research and drug discovery.

FunCoup 6: advancing functional association networks across species with directed links and improved user experience

FunCoup 6 (https://funcoup.org) represents a significant advancement in global functional association networks, aiming to provide researchers with a comprehensive view of the functional coupling interactome. This update introduces novel methodologies and integrated tools for improved network inference and analysis. Major new developments in FunCoup 6 include vastly expanding the coverage of gene regulatory links, a new framework for bin-free Bayesian training and a new website. FunCoup 6 integrates a new tool for disease and drug target module identification using the TOPAS algorithm. To expand the utility of the resource for biomedical research, it incorporates pathway enrichment analysis using the ANUBIX and EASE algorithms. The unique comparative interactomics analysis in FunCoup provides insights of network conservation, now allowing users to align orthologs only or query each species network independently. Bin-free training was applied to 23 primary species, and in addition, networks were generated for all remaining 618 species in InParanoiDB 9. Accompanying these advancements, FunCoup 6 features a new redesigned website, together with updated API functionalities, and represents a pivotal step forward in functional genomics research, offering unique capabilities for exploring the complex landscape of protein interactions.

State of the interactomes: an evaluation of molecular networks for generating biological insights

Advancements in genomic and proteomic technologies have powered the creation of large gene and protein networks ("interactomes") for understanding biological systems. However, the proliferation of interactomes complicates the selection of networks for specific applications. Here, we present a comprehensive evaluation of 45 current human interactomes, encompassing protein-protein interactions as well as gene regulatory, signaling, colocalization, and genetic interaction networks. Our analysis shows that large composite networks such as HumanNet, STRING, and FunCoup are most effective for identifying disease genes, while smaller networks such as DIP, Reactome, and SIGNOR demonstrate stronger performance in interaction prediction. Our study provides a benchmark for interactomes across diverse biological applications and clarifies factors that influence network performance. Furthermore, our evaluation pipeline paves the way for continued assessment of emerging and updated interaction networks in the future.

Dock2D: Synthetic Data for the Molecular Recognition Problem

Predicting the physical interaction of proteins is a cornerstone problem in computational biology. New classes of learning-based algorithms are actively being developed, and are typically trained end-to-end on protein complex structures extracted from the Protein Data Bank. These training datasets tend to be large and difficult to use for prototyping and, unlike image or natural language datasets, they are not easily interpretable by non-experts. We present Dock2D-IP and Dock2D-IF, two "toy" datasets that can be used to select algorithms predicting protein-protein interactions-or any other type of molecular interactions. Using two-dimensional shapes as input, each example from Dock2D-IP ("interaction pose") describes the interaction pose of two shapes known to interact and each example from Dock2D-IF ("interaction fact") describes whether two shapes form a stable complex or not, regardless of how they bind. We propose a number of baseline solutions to the problem and show that the same underlying energy function can be learned either by solving the interaction pose task (formulated as an energy-minimization "docking" problem) or the fact-of-interaction task (formulated as a binding free energy estimation problem).

Improved protein interaction models predict differences in complexes between human cell lines

The interactions of proteins to form complexes play a crucial role in cell function. Data on protein-protein or pairwise interactions (PPI) typically come from a combination of sample separation and mass spectrometry. Since 2010, several extensive, high-throughput mass spectrometry-based experimental studies have dramatically expanded public repositories for PPI data and, by extension, our knowledge of protein complexes. Unfortunately, challenges of limited overlap between experiments, modality-oriented biases, and prohibitive costs of experimental reproducibility continue to limit coverage of the human protein assembly map, both underscoring the need for and spurring the development of relevant computational approaches. Here, we present a new method for predicting the strength of protein interactions. It addresses two important issues that have limited past PPI prediction approaches: incomplete feature sets and incomplete proteome coverage. For a given collection of protein pairs, we fused data from heterogeneous sources into a feature matrix and identified the minimal set of feature partitions for which a non-empty set of protein pairs had complete values. For each such feature partition, we trained a classifier to predict PPI probabilities. We then calculated an overall prediction for a given protein pair by weighting the probabilities from all models that applied to that pair. Our approach accurately identified known and highly probable PPI, far exceeding the performance of current approaches and providing more complete proteome coverage. We then used the predicted probabilities to assemble complexes using previously-described graph-based tools and clustering algorithms and again obtained improved results. Lastly, we used features for three human cell lines to predict PPI and complex scores and identified complexes predicted to differ between those cell lines.

A structurally informed human protein-protein interactome reveals proteome-wide perturbations caused by disease mutations

To assist the translation of genetic findings to disease pathobiology and therapeutics discovery, we present an ensemble deep learning framework, termed PIONEER (Protein-protein InteractiOn iNtErfacE pRediction), that predicts protein-binding partner-specific interfaces for all known protein interactions in humans and seven other common model organisms to generate comprehensive structurally informed protein interactomes. We demonstrate that PIONEER outperforms existing state-of-the-art methods and experimentally validate its predictions. We show that disease-associated mutations are enriched in PIONEER-predicted protein-protein interfaces and explore their impact on disease prognosis and drug responses. We identify 586 significant protein-protein interactions (PPIs) enriched with PIONEER-predicted interface somatic mutations (termed oncoPPIs) from analysis of approximately 11,000 whole exomes across 33 cancer types and show significant associations of oncoPPIs with patient survival and drug responses. PIONEER, implemented as both a web server platform and a software package, identifies functional consequences of disease-associated alleles and offers a deep learning tool for precision medicine at multiscale interactome network levels.

MAGPIE: An interactive tool for visualizing and analyzing protein-ligand interactions

Quantitative tools to compile and analyze biomolecular interactions among chemically diverse binding partners would improve therapeutic design and aid in studying molecular evolution. Here we present Mapping Areas of Genetic Parsimony In Epitopes (MAGPIE), a publicly available software package for simultaneously visualizing and analyzing thousands of interactions between a single protein or small molecule ligand (the "target") and all of its protein binding partners ("binders"). MAGPIE generates an interactive three-dimensional visualization from a set of protein complex structures that share the target ligand, as well as sequence logo-style amino acid frequency graphs that show all the amino acids from the set of protein binders that interact with user-defined target ligand positions or chemical groups. MAGPIE highlights all the salt bridge and hydrogen bond interactions made by the target in the visualization and as separate amino acid frequency graphs. Finally, MAGPIE collates the most common target-binder interactions as a list of "hotspots," which can be used to analyze trends or guide the de novo design of protein binders. As an example of the utility of the program, we used MAGPIE to probe how different antibody fragments bind a viral antigen; how a common metabolite binds diverse protein partners; and how two ligands bind orthologs of a well-conserved glycolytic enzyme for a detailed understanding of evolutionarily conserved interactions involved in its activation and inhibition. MAGPIE is implemented in Python 3 and freely available at https://github.com/glasgowlab/MAGPIE, along with sample datasets, usage examples, and helper scripts to prepare input structures.

SLiMAn 2.0: meaningful navigation through peptide-protein interaction networks

Among the myriad of protein-protein interactions occurring in living organisms, a substantial amount involves small linear motifs (SLiMs) recognized by structured domains. However, predictions of SLiM-based networks are tedious, due to the abundance of such motifs and a high portion of false positive hits. For this reason, a webserver SLiMAn (Short Linear Motif Analysis) was developed to focus the search on the most relevant SLiMs. Using SLiMAn, one can navigate into a given (meta-)interactome and tune a variety of parameters associated to each type of SLiMs in attempt to identify functional ELM motifs and their recognition domains. The IntAct and BioGRID databases bring experimental information, while IUPred and AlphaFold provide boundaries of folded and disordered regions. Post-translational modifications listed in PhosphoSite+ are highlighted. Links to PubMed accelerate scrutiny into the literature, to support (or not) putative pairings. Dedicated visualization features are also incorporated, such as Cytoscape for macromolecular networks and BINANA for intermolecular contacts within structural models generated by SCWRL 3.0. The use of SLiMAn 2.0 is illustrated on a simple example. It is freely available at https://sliman2.cbs.cnrs.fr.

ZEPPI: Proteome-scale sequence-based evaluation of protein-protein interaction models

We introduce ZEPPI (Z-score Evaluation of Protein-Protein Interfaces), a framework to evaluate structural models of a complex based on sequence coevolution and conservation involving residues in protein-protein interfaces. The ZEPPI score is calculated by comparing metrics for an interface to those obtained from randomly chosen residues. Since contacting residues are defined by the structural model, this obviates the need to account for indirect interactions. Further, although ZEPPI relies on species-paired multiple sequence alignments, its focus on interfacial residues allows it to leverage quite shallow alignments. ZEPPI can be implemented on a proteome-wide scale and is applied here to millions of structural models of dimeric complexes in the Escherichia coli and human interactomes found in the PrePPI database. PrePPI's scoring function is based primarily on the evaluation of protein-protein interfaces, and ZEPPI adds a new feature to this analysis through the incorporation of evolutionary information. ZEPPI performance is evaluated through applications to experimentally determined complexes and to decoys from the CASP-CAPRI experiment. As we discuss, the standard CAPRI scores used to evaluate docking models are based on model quality and not on the ability to give yes/no answers as to whether two proteins interact. ZEPPI is able to detect weak signals from PPI models that the CAPRI scores define as incorrect and, similarly, to identify potential PPIs defined as low confidence by the current PrePPI scoring function. A number of examples that illustrate how the combination of PrePPI and ZEPPI can yield functional hypotheses are provided.

Prediction of interactions between cell surface proteins by machine learning

Cells detect changes in their external environments or communicate with each other through proteins on their surfaces. These cell surface proteins form a complicated network of interactions in order to fulfill their functions. The interactions between cell surface proteins are highly dynamic and, thus, challenging to detect using traditional experimental techniques. Here, we tackle this challenge using a computational framework. The primary focus of the framework is to develop new tools to identify interactions between domains in the immunoglobulin (Ig) fold, which is the most abundant domain family in cell surface proteins. These interactions could be formed between ligands and receptors from different cells or between proteins on the same cell surface. In practice, we collected all structural data on Ig domain interactions and transformed them into an interface fragment pair library. A high-dimensional profile can then be constructed from the library for a given pair of query protein sequences. Multiple machine learning models were used to read this profile so that the probability of interaction between the query proteins could be predicted. We tested our models on an experimentally derived dataset that contains 564 cell surface proteins in humans. The cross-validation results show that we can achieve higher than 70% accuracy in identifying the PPIs within this dataset. We then applied this method to a group of 46 cell surface proteins in Caenorhabditis elegans. We screened every possible interaction between these proteins. Many interactions recognized by our machine learning classifiers have been experimentally confirmed in the literature. In conclusion, our computational platform serves as a useful tool to help identify potential new interactions between cell surface proteins in addition to current state-of-the-art experimental techniques. The tool is freely accessible for use by the scientific community. Moreover, the general framework of the machine learning classification can also be extended to study the interactions of proteins in other domain superfamilies.

Revolutionizing protein-protein interaction prediction with deep learning

Protein-protein interactions (PPIs) are pivotal for driving diverse biological processes, and any disturbance in these interactions can lead to disease. Thus, the study of PPIs has been a central focus in biology. Recent developments in deep learning methods, coupled with the vast genomic sequence data, have significantly boosted the accuracy of predicting protein structures and modeling protein complexes, approaching levels comparable to experimental techniques. Herein, we review the latest advances in the computational methods for modeling 3D protein complexes and the prediction of protein interaction partners, emphasizing the application of deep learning methods deriving from coevolution analysis. The review also highlights biomedical applications of PPI prediction and outlines challenges in the field.

SMARCAL1 is a dual regulator of innate immune signaling and PD-L1 expression that promotes tumor immune evasion

Genomic instability can trigger cancer-intrinsic innate immune responses that promote tumor rejection. However, cancer cells often evade these responses by overexpressing immune checkpoint regulators, such as PD-L1. Here, we identify the SNF2-family DNA translocase SMARCAL1 as a factor that favors tumor immune evasion by a dual mechanism involving both the suppression of innate immune signaling and the induction of PD-L1-mediated immune checkpoint responses. Mechanistically, SMARCAL1 limits endogenous DNA damage, thereby suppressing cGAS-STING-dependent signaling during cancer cell growth. Simultaneously, it cooperates with the AP-1 family member JUN to maintain chromatin accessibility at a PD-L1 transcriptional regulatory element, thereby promoting PD-L1 expression in cancer cells. SMARCAL1 loss hinders the ability of tumor cells to induce PD-L1 in response to genomic instability, enhances anti-tumor immune responses and sensitizes tumors to immune checkpoint blockade in a mouse melanoma model. Collectively, these studies uncover SMARCAL1 as a promising target for cancer immunotherapy.

Prediction of influential proteins and enzymes of certain diseases using a directed unimodular hypergraph

Protein-protein interaction (PPI) analysis based on mathematical modeling is an efficient means of identifying hub proteins, corresponding enzymes and many underlying structures. In this paper, a method for the analysis of PPI is introduced and used to analyze protein interactions of diseases such as Parkinson's, COVID-19 and diabetes melitus. A directed hypergraph is used to represent PPI interactions. A novel directed hypergraph depth-first search algorithm is introduced to find the longest paths. The minor hypergraph reduces the dimension of the directed hypergraph, representing the longest paths and results in the unimodular hypergraph. The property of unimodular hypergraph clusters influential proteins and enzymes that are related thereby providing potential avenues for disease treatment.

What Strengthens Protein-Protein Interactions: Analysis and Applications of Residue Correlation Networks

Identifying residues critical to protein-protein binding and efficient design of stable and specific protein binders are challenging tasks. Extending beyond the direct contacts in a protein-protein binding interface, our study employs computational modeling to reveal the essential network of residue interactions and dihedral angle correlations critical in protein-protein recognition. We hypothesized that mutating residues exhibiting highly correlated dynamic motion within the interaction network could efficiently optimize protein-protein interactions to create tight and selective protein binders. We tested this hypothesis using the ubiquitin (Ub) and MERS coronaviral papain-like protease (PLpro) complex, since Ub is a central player in multiple cellular functions and PLpro is an antiviral drug target. Our designed ubiquitin variant (UbV) hosting three mutated residues displayed a ∼3,500-fold increase in functional inhibition relative to wild-type Ub. Further optimization of two C-terminal residues within the Ub network resulted in a KD of 1.5 nM and IC50 of 9.7 nM for the five-point Ub mutant, eliciting 27,500-fold and 5,500-fold enhancements in affinity and potency, respectively, as well as improved selectivity, without destabilizing the UbV structure. Our study highlights residue correlation and interaction networks in protein-protein interactions, and introduces an effective approach to design high-affinity protein binders for cell biology research and future therapeutics.

ZEPPI: proteome-scale sequence-based evaluation of protein-protein interaction models

We introduce ZEPPI (Z-score Evaluation of Protein-Protein Interfaces), a framework to evaluate structural models of a complex based on sequence co-evolution and conservation involving residues in protein-protein interfaces. The ZEPPI score is calculated by comparing metrics for an interface to those obtained from randomly chosen residues. Since contacting residues are defined by the structural model, this obviates the need to account for indirect interactions. Further, although ZEPPI relies on species-paired multiple sequence alignments, its focus on interfacial residues allows it to leverage quite shallow alignments. ZEPPI performance is evaluated through applications to experimentally determined complexes and to decoys from the CASP-CAPRI experiment. ZEPPI can be implemented on a proteome-wide scale as evidenced by calculations on millions of structural models of dimeric complexes in the E. coli and human interactomes found in the PrePPI database. A number of examples that illustrate how these tools can yield novel functional hypotheses are provided.

Elucidation of genome-wide understudied proteins targeted by PROTAC-induced degradation using interpretable machine learning

Proteolysis-targeting chimeras (PROTACs) are hetero-bifunctional molecules that induce the degradation of target proteins by recruiting an E3 ligase. PROTACs have the potential to inactivate disease-related genes that are considered undruggable by small molecules, making them a promising therapy for the treatment of incurable diseases. However, only a few hundred proteins have been experimentally tested for their amenability to PROTACs, and it remains unclear which other proteins in the entire human genome can be targeted by PROTACs. In this study, we have developed PrePROTAC, an interpretable machine learning model based on a transformer-based protein sequence descriptor and random forest classification. PrePROTAC predicts genome-wide targets that can be degraded by CRBN, one of the E3 ligases. In the benchmark studies, PrePROTAC achieved a ROC-AUC of 0.81, an average precision of 0.84, and over 40% sensitivity at a false positive rate of 0.05. When evaluated by an external test set which comprised proteins from different structural folds than those in the training set, the performance of PrePROTAC did not drop significantly, indicating its generalizability. Furthermore, we developed an embedding SHapley Additive exPlanations (eSHAP) method, which extends conventional SHAP analysis for original features to an embedding space through in silico mutagenesis. This method allowed us to identify key residues in the protein structure that play critical roles in PROTAC activity. The identified key residues were consistent with existing knowledge. Using PrePROTAC, we identified over 600 novel understudied proteins that are potentially degradable by CRBN and proposed PROTAC compounds for three novel drug targets associated with Alzheimer's disease.

Identification of Protein Complexes by Integrating Protein Abundance and Interaction Features Using a Deep Learning Strategy

Many essential cellular functions are carried out by multi-protein complexes that can be characterized by their protein-protein interactions. The interactions between protein subunits are critically dependent on the strengths of their interactions and their cellular abundances, both of which span orders of magnitude. Despite many efforts devoted to the global discovery of protein complexes by integrating large-scale protein abundance and interaction features, there is still room for improvement. Here, we integrated >7000 quantitative proteomic samples with three published affinity purification/co-fractionation mass spectrometry datasets into a deep learning framework to predict protein-protein interactions (PPIs), followed by the identification of protein complexes using a two-stage clustering strategy. Our deep-learning-technique-based classifier significantly outperformed recently published machine learning prediction models and in the process captured 5010 complexes containing over 9000 unique proteins. The vast majority of proteins in our predicted complexes exhibited low or no tissue specificity, which is an indication that the observed complexes tend to be ubiquitously expressed throughout all cell types and tissues. Interestingly, our combined approach increased the model sensitivity for low abundant proteins, which amongst other things allowed us to detect the interaction of MCM10, which connects to the replicative helicase complex via the MCM6 protein. The integration of protein abundances and their interaction features using a deep learning approach provided a comprehensive map of protein-protein interactions and a unique perspective on possible novel protein complexes.

Growing ecosystem of deep learning methods for modeling protein-protein interactions

Numerous cellular functions rely on protein-protein interactions. Efforts to comprehensively characterize them remain challenged however by the diversity of molecular recognition mechanisms employed within the proteome. Deep learning has emerged as a promising approach for tackling this problem by exploiting both experimental data and basic biophysical knowledge about protein interactions. Here, we review the growing ecosystem of deep learning methods for modeling protein interactions, highlighting the diversity of these biophysically informed models and their respective trade-offs. We discuss recent successes in using representation learning to capture complex features pertinent to predicting protein interactions and interaction sites, geometric deep learning to reason over protein structures and predict complex structures, and generative modeling to design de novo protein assemblies. We also outline some of the outstanding challenges and promising new directions. Opportunities abound to discover novel interactions, elucidate their physical mechanisms, and engineer binders to modulate their functions using deep learning and, ultimately, unravel how protein interactions orchestrate complex cellular behaviors.