Protein-protein interactions: from global to local analyses

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.

Information processing by simple molecular motifs and susceptibility to noise

Biological organisms rely on their ability to sense and respond appropriately to their environment. The molecular mechanisms that facilitate these essential processes are however subject to a range of random effects and stochastic processes, which jointly affect the reliability of information transmission between receptors and, for example, the physiological downstream response. Information is mathematically defined in terms of the entropy; and the extent of information flowing across an information channel or signalling system is typically measured by the 'mutual information', or the reduction in the uncertainty about the output once the input signal is known. Here, we quantify how extrinsic and intrinsic noise affects the transmission of simple signals along simple motifs of molecular interaction networks. Even for very simple systems, the effects of the different sources of variability alone and in combination can give rise to bewildering complexity. In particular, extrinsic variability is apt to generate 'apparent' information that can, in extreme cases, mask the actual information that for a single system would flow between the different molecular components making up cellular signalling pathways. We show how this artificial inflation in apparent information arises and how the effects of different types of noise alone and in combination can be understood.

Nataraja K. An Approach to Function Annotation for Proteins of Unknown Function (PUFs) in the Transcriptome of Indian Mulberry

The modern sequencing technologies are generating large volumes of information at the transcriptome and genome level. Translation of this information into a biological meaning is far behind the race due to which a significant portion of proteins discovered remain as proteins of unknown function (PUFs). Attempts to uncover the functional significance of PUFs are limited due to lack of easy and high throughput functional annotation tools. Here, we report an approach to assign putative functions to PUFs, identified in the transcriptome of mulberry, a perennial tree commonly cultivated as host of silkworm. We utilized the mulberry PUFs generated from leaf tissues exposed to drought stress at whole plant level. A sequence and structure based computational analysis predicted the probable function of the PUFs. For rapid and easy annotation of PUFs, we developed an automated pipeline by integrating diverse bioinformatics tools, designated as PUFs Annotation Server (PUFAS), which also provides a web service API (Application Programming Interface) for a large-scale analysis up to a genome. The expression analysis of three selected PUFs annotated by the pipeline revealed abiotic stress responsiveness of the genes, and hence their potential role in stress acclimation pathways. The automated pipeline developed here could be extended to assign functions to PUFs from any organism in general. PUFAS web server is available at http://caps.ncbs.res.in/pufas/ and the web service is accessible at http://capservices.ncbs.res.in/help/pufas.

The role of protein interaction networks in systems biomedicine

The challenging task of studying and modeling complex dynamics of biological systems in order to describe various human diseases has gathered great interest in recent years. Major biological processes are mediated through protein interactions, hence there is a need to understand the chaotic network that forms these processes in pursuance of understanding human diseases. The applications of protein interaction networks to disease datasets allow the identification of genes and proteins associated with diseases, the study of network properties, identification of subnetworks, and network-based disease gene classification. Although various protein interaction network analysis strategies have been employed, grand challenges are still existing. Global understanding of protein interaction networks via integration of high-throughput functional genomics data from different levels will allow researchers to examine the disease pathways and identify strategies to control them. As a result, it seems likely that more personalized, more accurate and more rapid disease gene diagnostic techniques will be devised in the future, as well as novel strategies that are more personalized. This mini-review summarizes the current practice of protein interaction networks in medical research as well as challenges to be overcome.

Proteome-wide protein interaction measurements of bacterial proteins of unknown function

Despite the enormous proliferation of bacterial genome data, surprisingly persistent collections of bacterial proteins have resisted functional annotation. In a typical genome, roughly 30% of genes have no assigned function. Many of these proteins are conserved across a large number of bacterial genomes. To assign a putative function to these conserved proteins of unknown function, we created a physical interaction map by measuring biophysical interaction of these proteins. Binary protein--protein interactions in the model organism Streptococcus pneumoniae (TIGR4) are measured with a microfluidic high-throughput assay technology. In some cases, informatic analysis was used to restrict the space of potential binding partners. In other cases, we performed in vitro proteome-wide interaction screens. We were able to assign putative functions to 50 conserved proteins of unknown function that we studied with this approach.

Proteomics in evolutionary ecology: linking the genotype with the phenotype

The study of the proteome (proteomics), which includes the dynamics of protein expression, regulation, interactions and its function, has played a less prominent role in evolutionary and ecological investigations in comparison with the study of the genome and transcriptome. There are, however, a number of arguments suggesting that this situation should change. First, the proteome is closer to the phenotype than the genome or the transcriptome, and as such may be more directly responsive to natural selection, and thus closely linked to adaptation. Second, there is evidence of a low correlation between protein and transcript expression levels across genes in many different organisms. Finally, there have been some recent important technological improvements in proteomics methods that make them feasible, practical and useful to address a wide range of evolutionary questions even in nonmodel organisms. The different proteomic methods, their limitations and problems when interpreting empirical data are described and discussed. In addition, the proteomic literature pertaining to evolutionary ecology is reviewed with examples, and potential applications of proteomics in a variety of evolutionary contexts are outlined. New proteomic research trends such as the study of posttranslational modifications and protein-protein interactions, as well as the combined use of the different -omics approaches, are discussed in relation to the development of a more functional and integrated perspective, needed for achieving a more comprehensive knowledge of evolutionary change.

Partner-aware prediction of interacting residues in protein-protein complexes from sequence data

Computational prediction of residues that participate in protein-protein interactions is a difficult task, and state of the art methods have shown only limited success in this arena. One possible problem with these methods is that they try to predict interacting residues without incorporating information about the partner protein, although it is unclear how much partner information could enhance prediction performance. To address this issue, the two following comparisons are of crucial significance: (a) comparison between the predictability of inter-protein residue pairs, i.e., predicting exactly which residue pairs interact with each other given two protein sequences; this can be achieved by either combining conventional single-protein predictions or making predictions using a new model trained directly on the residue pairs, and the performance of these two approaches may be compared: (b) comparison between the predictability of the interacting residues in a single protein (irrespective of the partner residue or protein) from conventional methods and predictions converted from the pair-wise trained model. Using these two streams of training and validation procedures and employing similar two-stage neural networks, we showed that the models trained on pair-wise contacts outperformed the partner-unaware models in predicting both interacting pairs and interacting single-protein residues. Prediction performance decreased with the size of the conformational change upon complex formation; this trend is similar to docking, even though no structural information was used in our prediction. An example application that predicts two partner-specific interfaces of a protein was shown to be effective, highlighting the potential of the proposed approach. Finally, a preliminary attempt was made to score docking decoy poses using prediction of interacting residue pairs; this analysis produced an encouraging result.

Assessing coverage of protein interaction data using capture-recapture models

Protein interaction networks comprise thousands of individual binary links between distinct proteins. Whilst these data have attracted considerable attention and been the focus of many different studies, the networks, their structure, function, and how they change over time are still not fully known. More importantly, there is still considerable uncertainty regarding their size, and the quality of the available data continues to be questioned. Here, we employ statistical models of the experimental sampling process, in particular capture-recapture methods, in order to assess the false discovery rate and size of protein interaction networks. We uses these methods to gauge the ability of different experimental systems to find the true binary interactome. Our model allows us to obtain estimates for the size and false-discovery rate from simple considerations regarding the number of repeatedly interactions, and provides suggestions as to how we can exploit this information in order to reduce the effects of noise in such data. In particular our approach does not require a reference dataset. We estimate that approximately more than half of the true physical interactome has now been sampled in yeast.

Protein-protein interactions: principles, techniques, and their potential role in new drug development

A vast network of genes is inter-linked through protein-protein interactions and is critical component of almost every biological process under physiological conditions. Any disruption of the biologically essential network leads to pathological conditions resulting into related diseases. Therefore, proper understanding of biological functions warrants a comprehensive knowledge of protein-protein interactions and the molecular mechanisms that govern such processes. The importance of protein-protein interaction process is highlighted by the fact that a number of powerful techniques/methods have been developed to understand how such interactions take place under various physiological and pathological conditions. Many of the key protein-protein interactions are known to participate in disease-associated signaling pathways, and represent novel targets for therapeutic intervention. Thus, controlling protein-protein interactions offers a rich dividend for the discovery of new drug targets. Availability of various tools to study and the knowledge of human genome have put us in a unique position to understand highly complex biological network, and the mechanisms involved therein. In this review article, we have summarized protein-protein interaction networks, techniques/methods of their binding/kinetic parameters, and the role of these interactions in the development of potential tools for drug designing.

Electron-nuclear interactions as probes of domain motion in proteins

Long range interactions between nuclear spins and paramagnetic ions can serve as a sensitive monitor of internal motion of various parts of proteins, including functional loops and separate domains. In the case of interdomain motion, the interactions between the ion and NMR-observable nuclei are modulated in direction and magnitude mainly by a combination of overall and interdomain motions. The effects on observable parameters such as paramagnetic relaxation enhancement (PRE) and pseudocontact shift (PCS) can, in principle, be used to characterize motion. These parameters are frequently used for the purpose of structural refinements. However, their use to probe actual domain motions is less common and is lacking a proper theoretical treatment from a motional perspective. In this work, a suitable spin Hamiltonian is incorporated in a two body diffusion model to produce the time correlation function for the nuclear spin-paramagnetic ion interactions. Simulated observables for nuclei in different positions with respect to the paramagnetic ion are produced. Based on these simulations, it demonstrated that both the PRE and the PCS can be very sensitive probes of domain motion. Results for different nuclei within the protein sense different aspects of the motions. Some are more sensitive to the amplitude of the internal motion, others are more sensitive to overall diffusion rates, allowing separation of these contributions. Experimentally, the interaction strength can also be tuned by substitution of different paramagnetic ions or by varying magnetic field strength (in the case of lanthanides) to allow the use of more detailed diffusion models without reducing the reliability of data fitting.

Efficient identification of tubby-binding proteins by an improved system of T7 phage display

Mutation in the tubby gene causes adult-onset obesity, progressive retinal, and cochlear degeneration with unknown mechanism. In contrast, mutations in tubby-like protein 1 (Tulp1), whose C-terminus is highly homologous to tubby, only lead to retinal degeneration. We speculate that their diverse N-terminus may define their distinct disease profile. To elucidate the binding partners of tubby, we used tubby N-terminus (tubby-N) as bait to identify unknown binding proteins with open-reading-frame (ORF) phage display. T7 phage display was engineered with three improvements: high-quality ORF phage display cDNA library, specific phage elution by protease cleavage, and dual phage display for sensitive high throughput screening. The new system is capable of identifying unknown bait-binding proteins in as fast as approximately 4-7 days. While phage display with conventional cDNA libraries identifies high percentage of out-of-frame unnatural short peptides, all 28 tubby-N-binding clones identified by ORF phage display were ORFs. They encode 16 proteins, including 8 nuclear proteins. Fourteen proteins were analyzed by yeast two-hybrid assay and protein pull-down assay with ten of them independently verified. Comparative binding analyses revealed several proteins binding to both tubby and Tulp1 as well as one tubby-specific binding protein. These data suggest that tubby-N is capable of interacting with multiple nuclear and cytoplasmic protein binding partners. These results demonstrated that the newly-engineered ORF phage display is a powerful technology to identify unknown protein-protein interactions.

An extensive survey of tyrosine phosphorylation revealing new sites in human mammary epithelial cells

Protein tyrosine phosphorylation represents a central regulatory mechanism in cell signaling. Here, we present an extensive survey of tyrosine phosphorylation sites in a normal-derived human mammary epithelial cell (HMEC) line by applying antiphosphotyrosine peptide immunoaffinity purification coupled with high sensitivity capillary liquid chromatography tandem mass spectrometry. A total of 481 tyrosine phosphorylation sites (covered by 716 unique peptides) from 285 proteins were confidently identified in HMEC following the analysis of both the basal condition and acute stimulation with epidermal growth factor (EGF). The estimated false discovery rate was 1.0% as determined by searching against a scrambled database. Comparison of these data with existing literature showed significant agreement for previously reported sites. However, we observed 281 sites that were not previously reported for HMEC cultures and 29 of which have not been reported for any human cell or tissue system. The analysis showed that a majority of highly phosphorylated proteins were relatively low-abundance. Large differences in phosphorylation stoichiometry for sites within the same protein were also observed, raising the possibility of more important functional roles for such highly phosphorylated pTyr sites. By mapping to major signaling networks, such as the EGF receptor and insulin growth factor-1 receptor signaling pathways, many known proteins involved in these pathways were revealed to be tyrosine phosphorylated, which provides interesting targets for future hypothesis-driven and targeted quantitative studies involving tyrosine phosphorylation in HMEC or other human systems.

Bioinformatics in bioinorganic chemistry

Bioinformatics is a central discipline in modern life sciences aimed at describing the complex properties of living organisms starting from large-scale data sets of cellular constituents such as genes and proteins. In order for this wealth of information to provide useful biological knowledge, databases and software tools for data collection, analysis and interpretation need to be developed. In this paper, we review recent advances in the design and implementation of bioinformatics resources devoted to the study of metals in biological systems, a research field traditionally at the heart of bioinorganic chemistry. We show how metalloproteomes can be extracted from genome sequences, how structural properties can be related to function, how databases can be implemented, and how hints on interactions can be obtained from bioinformatics.

The modern RNP world of eukaryotes

Eukaryote gene expression is mediated by a cascade of RNA functions that regulate, process, store, transport, and translate RNA transcripts. The RNA network that promotes this cascade depends on a large cohort of proteins that partner RNAs; thus, the modern RNA world of eukaryotes is really a ribonucleoprotein (RNP) world. Features of this "RNP infrastructure" can be related to the high cytosolic density of macromolecules and the large size of eukaryote cells. Because of the densely packed cytosol or nucleoplasm (with its severe restriction on diffusion of macromolecules), partitioning of the eukaryote cell into functionally specialized compartments is essential for efficiency. This necessitates the association of RNA and protein into large RNP complexes including ribosomes and spliceosomes. This is well illustrated by the ubiquitous spliceosome for which most components are conserved throughout eukaryotes and which interacts with other RNP-based machineries. The complexes involved in gene processing in modern eukaryotes have broad phylogenetic distributions suggesting that the common ancestor of extant eukaryotes had a fully evolved RNP network. Thus, the eukaryote genome may be uniquely informative about the transition from an earlier RNA genome world to the modern DNA genome world.

Yeast two-hybrid, a powerful tool for systems biology

A key property of complex biological systems is the presence of interaction networks formed by its different components, primarily proteins. These are crucial for all levels of cellular function, including architecture, metabolism and signalling, as well as the availability of cellular energy. Very stable, but also rather transient and dynamic protein-protein interactions generate new system properties at the level of multiprotein complexes, cellular compartments or the entire cell. Thus, interactomics is expected to largely contribute to emerging fields like systems biology or systems bioenergetics. The more recent technological development of high-throughput methods for interactomics research will dramatically increase our knowledge of protein interaction networks. The two most frequently used methods are yeast two-hybrid (Y2H) screening, a well established genetic in vivo approach, and affinity purification of complexes followed by mass spectrometry analysis, an emerging biochemical in vitro technique. So far, a majority of published interactions have been detected using an Y2H screen. However, with the massive application of this method, also some limitations have become apparent. This review provides an overview on available yeast two-hybrid methods, in particular focusing on more recent approaches. These allow detection of protein interactions in their native environment, as e.g. in the cytosol or bound to a membrane, by using cytosolic signalling cascades or split protein constructs. Strengths and weaknesses of these genetic methods are discussed and some guidelines for verification of detected protein-protein interactions are provided.

Accurate characterization of weak macromolecular interactions by titration of NMR residual dipolar couplings: application to the CD2AP SH3-C:ubiquitin complex

The description of the interactome represents one of key challenges remaining for structural biology. Physiologically important weak interactions, with dissociation constants above 100 μM, are remarkably common, but remain beyond the reach of most of structural biology. NMR spectroscopy, and in particular, residual dipolar couplings (RDCs) provide crucial conformational constraints on intermolecular orientation in molecular complexes, but the combination of free and bound contributions to the measured RDC seriously complicates their exploitation for weakly interacting partners. We develop a robust approach for the determination of weak complexes based on: (i) differential isotopic labeling of the partner proteins facilitating RDC measurement in both partners; (ii) measurement of RDC changes upon titration into different equilibrium mixtures of partially aligned free and complex forms of the proteins; (iii) novel analytical approaches to determine the effective alignment in all equilibrium mixtures; and (iv) extraction of precise RDCs for bound forms of both partner proteins. The approach is demonstrated for the determination of the three-dimensional structure of the weakly interacting CD2AP SH3-C:Ubiquitin complex (K_d = 132 ± 13 μM) and is shown, using cross-validation, to be highly precise. We expect this methodology to extend the remarkable and unique ability of NMR to study weak protein–protein complexes.