Weak Protein-protein Interactions Research Articles

Proximal Co-Translation Facilitates Detection of Weak Protein-Protein Interactions

Ubiquitin (Ub) signals are recognized and decoded into cellular responses by Ub-receptors, proteins that tether the Ub-binding domain(s) (UBDs) with response elements. Typically, UBDs bind mono-Ub in highly dynamic and weak affinity manners, presenting challenges in identifying and characterizing their binding interfaces. Here, we report the development of a new approach to facilitate the detection of these weak interactions using split-reporter systems where two interacting proteins are proximally co-translated from a single mRNA. This proximity significantly enhances the readout signals of weak protein–protein interactions (PPIs). We harnessed this system to characterize the ultra-weak UBD and ENTH (Epsin N-terminal Homology) and discovered that the yeast Ent1-ENTH domain contains two Ub-binding patches. One is similar to a previously characterized patch on STAM1(signal-transducing adaptor molecule)-VHS (Vps27, Hrs, and STAM), and the other was predicted by AlphaFold. Using a split-CAT selection system that co-translates Ub and ENTH in combination with mutagenesis, we assessed and confirmed the existence of a novel binding patch around residue F53 on ENTH. Co-translation in the split-CAT system provides an effective tool for studying weak PPIs and offers new insights into Ub-receptor interactions.

High-throughput affinity measurements of direct interactions between activation domains and co-activators.

Sequence-specific activation by transcription factors is essential for gene regulation1,2. Key to this are activation domains, which often fall within disordered regions of transcription factors3,4 and recruit co-activators to initiate transcription5. These interactions are difficult to characterize via most experimental techniques because they are typically weak and transient6,7. Consequently, we know very little about whether these interactions are promiscuous or specific, the mechanisms of binding, and how these interactions tune the strength of gene activation. To address these questions, we developed a microfluidic platform for expression and purification of hundreds of activation domains in parallel followed by direct measurement of co-activator binding affinities (STAMMPPING, for Simultaneous Trapping of Affinity Measurements via a Microfluidic Protein-Protein INteraction Generator). By applying STAMMPPING to quantify direct interactions between eight co-activators and 204 human activation domains (>1,500 K ds), we provide the first quantitative map of these interactions and reveal 334 novel binding pairs. We find that the metazoan-specific co-activator P300 directly binds >100 activation domains, potentially explaining its widespread recruitment across the genome to influence transcriptional activation. Despite sharing similar molecular properties (e.g. enrichment of negative and hydrophobic residues), activation domains utilize distinct biophysical properties to recruit certain co-activator domains. Co-activator domain affinity and occupancy are well-predicted by analytical models that account for multivalency, and in vitro affinities quantitatively predict activation in cells with an ultrasensitive response. Not only do our results demonstrate the ability to measure affinities between even weak protein-protein interactions in high throughput, but they also provide a necessary resource of over 1,500 activation domain/co-activator affinities which lays the foundation for understanding the molecular basis of transcriptional activation.

Modulating Weak Protein-Protein Cross-Interactions by the Addition of Free Amino Acids at Millimolar Concentrations.

In this paper, we quantify weak protein-protein interactions in solution using cross-interaction chromatography (CIC) and surface plasmon resonance (SPR) and demonstrate that they can be modulated by the addition of millimolar concentrations of free amino acids. With CIC, we determined the second osmotic virial cross-interaction coefficient (B23) as a proxy for the interaction strength between two different proteins. We perform SPR experiments to establish the binding affinity between the same proteins. With CIC, we show that the amino acids proline, glutamine, and arginine render the protein cross-interactions more repulsive or equivalently less attractive. Specifically, we measured B23 between lysozyme (Lys) and bovine serum albumin (BSA) and between Lys and protein isolates (whey and canola). We find that B23 increases when amino acids are added to the solution even at millimolar concentrations, corresponding to protein/ligand stoichiometric ratios as low as 1:1. With SPR, we show that the binding affinity between proteins can change by 1 order of magnitude when 10 mM glutamine is added. In the case of Lys and one whey protein isolate (WPI), it changes from the mM to the M range, thus by 3 orders of magnitude. Interestingly, this efficient modulation of the protein cross-interactions does not alter the protein's secondary structure. The capacity of amino acids to modulate protein cross-interactions at mM concentrations is remarkable and may have an impact across fields in particular for specific applications in the food or pharmaceutical industries.

Advances in crosslinking chemistry and proximity-enabled strategies: deciphering protein complexes and interactions.

Mass spectrometry, coupled with innovative crosslinking techniques to decode protein conformations and interactions through uninterrupted signal connections, has undergone remarkable progress in recent years. It is crucial to develop selective crosslinking reagents that minimally disrupt protein structure and dynamics, providing insights into protein network regulation and biological functions. Compared to traditional crosslinkers, new bifunctional chemical crosslinkers exhibit high selectivity and specificity in connecting proximal amino acid residues, resulting in stable molecular crosslinked products. The conjugation with specific amino acid residues like lysine, cysteine, arginine and tyrosine expands the XL-MS toolbox, enabling more precise modeling of target substrates and leading to improved data quality and reliability. Another emerging crosslinking method utilizes unnatural amino acids (UAAs) derived from proximity-enabled reactivity with specific amino acids or sulfur-fluoride exchange (SuFEx) reactions with nucleophilic residues. These UAAs are genetically encoded into proteins for the formation of specific covalent bonds. This technique combines the benefits of genetic encoding for live cell compatibility with chemical crosslinking, providing a valuable method for capturing transient and weak protein-protein interactions (PPIs) for mapping PPI coordinates and improving the pharmacological properties of proteins. With continued advancements in technology and applications, crosslinking mass spectrometry is poised to play an increasingly significant role in guiding our understanding of protein dynamics and function in the future.

Methyl and Fluorine Effects in Novel Orally Bioavailable Keap1-Nrf2 PPI Inhibitor.

Oxidative stress is one of the causes of progression of chronic kidney disease (CKD). Activation of the antioxidant protein regulator Nrf2 by inhibition of the Keap1-Nrf2 protein-protein interaction (PPI) is of interest as a potential treatment for CKD. We report the identification of the novel and weak PPI inhibitor 7 with good physical properties by a high throughput screening (HTS) campaign, followed by structural and computational analysis. The installation of only methyl and fluorine groups successfully provided the lead compound 25, which showed more than 400-fold stronger activity. Furthermore, these dramatic substituent effects can be explained by the analysis of using isothermal titration calorimetry (ITC). Thus, the resulting 25, which exhibited high oral absorption and durability, would be a CKD therapeutic agent because of the dose-dependent manner for up-regulation of the antioxidant protein heme oxigenase-1 (HO-1) in rat kidneys.

Effects of Interfacial Interactions on Electrocatalytic Activity of Cytochrome c Oxidase in Biomimetic Lipid Membranes on Gold Electrodes.

Effects of interfacial interactions on the electrocatalytic activity of protein-tethered bilayer lipid membranes (ptBLMs) containing cytochrome c oxidase (CcO) for the oxygen reduction reaction are studied by using protein film electrochemistry and surface-enhanced infrared absorption (SEIRA) spectroscopy. Mammalian CcO was immobilized on a gold electrode via self-assembled monolayers (SAMs) of mixed alkanethiols. The protein orientation on the electrode is controlled by SAM-CcO interactions and is critical to the cytochrome c (cyt c) binding. The CcO-phospholipid and CcO-cyt c interactions modulate the electrocatalytic activity of CcO, and more densely packed ptBLMs show higher electrocatalytic activity. Our study indicates that spectroscopic and electrochemical studies of ptBLMs can provide insights into the effects of relatively weak protein-protein and protein-lipid interactions on the enzymatic activity of transmembrane enzymes.

Suppressing Nonspecific Binding in Biolayer Interferometry Experiments for Weak Ligand-Analyte Interactions.

Quantitative analysis of protein–protein interactions (PPIs) using biolayer interferometry (BLI) requires effective suppression of nonspecific binding (NSB) between analytes and biosensors. In particular, the study of weak interactions (i.e., KD > 1 μM) requires high concentrations of analytes, which substantially increases NSB. However, there are only a few so-called NSB blockers compatible with biomolecules, which limits the use of BLI in the accurate analysis of weak interactions. The present study aims to identify a new NSB blocker for the quantitative analysis of weak PPIs using BLI. We find that saccharides, especially sucrose, are potent NSB blockers and demonstrate their compatibility with other blocking additives. We also demonstrate the effects of the new NSB blocker by characterizing the binding between nonstructural protein 1 of the influenza A virus and human phosphoinositide 3-kinase. We anticipate that the new NSB-blocking admixture will find broad applications in studying weak interactions using BLI.

Characterization of protein corona formation on nanoparticles via the analysis of dynamic interfacial properties: Bovine serum albumin - silica particle interaction

Protein corona adsorption layers on nanoparticle surfaces, dispersed in biological fluids, can significantly change the interfacial interactions, reactivity, and mobility of the original nanoparticles designed as a drug carrier or existing as bioaerosols, bacteria, or viruses. The evaluation of the level of interactions (hard/soft corona), dispersion stability, and the ratio of the attached proteins per nanoparticle are essential parameters for the characterization of the protein corona formation on nanoparticle (PCN). In spite of development of several experimental techniques for this purpose, still more powerful, economical and fast measuring techniques are needed to work in-situ at original solution samples (near the physiological conditions), without requirement of additional sample preparation which can change the quality/quantity of the original interactions. In this study, a novel experimental protocol based on the analysis of dynamic interfacial properties (ADIP) is developed for in-situ evaluation of the protein–nanoparticle interactions under original conditions. For this purpose, dynamic surface tension and interfacial elasticity values of the bovine serum albumin (BSA) solutions alone and in mixed solutions with silica nanofluids are measured using drop profile analysis tensiometry. A considerable difference between the dynamic surface tension of BSA and protein corona solutions (BSA + SiO2 NPs complexes) demonstrates significant adsorption of the protein molecules at nanoparticles, confirmed by Fourier Transform Infrared Spectroscopy (FTIR) analysis. PCN complexes illustrate much slower kinetics of adsorption due to a smaller diffusion coefficient according to their larger size. The free proteins available in the solution were estimated considering early-time values of the dynamic surface tension, used for estimation of the adsorbed proteins per unit area of the silica surface (mol/cm2), considering the initial protein concentration in the bulk. The results show very good agreement with others’ results provided by AFM, DLS, UV–vis Spectroscopy, Multi-Parametric Surface Plasmon Resonance (MP-SPR), and Quartz Crystal Microbalance (QCM). Our novel experimental protocols and data analysis, demonstrates promising results for differentiation of the hard and soft corona layers, and unique for soft corona layer recognition, which is difficult by other available techniques, due to quick disturbances of weak protein-protein interaction in soft corona. The measured interfacial elasticity values confirm PCN formation and provides additional information for better differentiations of the corona layers.

Fast cross-linking by DOPA2 promotes the capturing of a stereospecific protein complex over nonspecific encounter complexes.

Transient and weak protein-protein interactions are essential to many biochemical reactions, yet are technically challenging to study. Chemical cross-linking of proteins coupled with mass spectrometry analysis (CXMS) provides a powerful tool in the analysis of such interactions. Central to this technology are chemical cross-linkers. Here, using two transient heterodimeric complexes EIN/HPr and EIIAGlc/EIIBGlc as our model systems, we evaluated the effects of two amine-specific homo-bifunctional cross-linkers with different reactivities. We showed previously that DOPA2 (di-ortho-phthalaldehyde with a di-ethylene glycol spacer arm) cross-links proteins 60-120 times faster than DSS (disuccinimidyl suberate). We found that though most of the intermolecular cross-links of either cross-linker are consistent with the encounter complexes (ECs), an ensemble of short-lived binding intermediates, more DOPA2 intermolecular cross-links could be assigned to the stereospecific complex (SC), the final lowest-energy conformational state for the two interacting proteins. Our finding suggests that faster cross-linking captures the SC more effectively and cross-linkers of different reactivities potentially probe protein-protein interaction dynamics across multiple timescales.

Dihydrocapsaicin induces translational repression and stress granule through HRI-eIF2α phosphorylation axis

Stress granules (SGs) are cytoplasmic biomolecular condensates that are formed against a variety of stress conditions when translation initiation is perturbed. SGs form through the weak protein-protein, protein-RNA, and RNA-RNA interactions, as well as through the intrinsically disordered domains and post-translation modifications within RNA binding proteins (RBPs). SGs are known to contribute to cell survivability by minimizing the stress-induced damage to the cells by delaying the activation of apoptosis. Here, we find that dihydrocapsaicin (DHC), an analogue of capsaicin, is a SG inducer that promotes polysome disassembly and reduces global protein translation via phosphorylation of eIF2α. DHC-mediated SG assembly is controlled by the phosphorylation of eIF2α at serine 51 position and is controlled by all four eIF2α stress kinases (i.e., HRI, PKR, PERK, and GCN2) with HRI showing maximal effect. We demonstrate that DHC is a bonafide compound that induces SG assembly, disassembles polysome, phosphorylates eIF2α in an HRI dependent manner, and thereby arrest global translation. Together, our results suggest that DHC is a novel SG inducer and an alternate to sodium arsenite to study SG dynamics.

Mapping the Proximity Interaction Network of STIM1 Reveals New Mechanisms of Cytoskeletal Regulation.

Stromal interaction molecule 1 (STIM1) resides primarily in the sarco/endoplasmic reticulum, where it senses intraluminal Ca2+ levels and activates Orai channels on the plasma membrane to initiate Ca2+ influx. We have previously shown that STIM1 is involved in the dynamic remodeling of the actin cytoskeleton. However, the downstream effectors of STIM1 that lead to cytoskeletal remodeling are not known. The proximity-labeling technique (BioID) can capture weak and transient protein-protein interactions, including proteins that reside in the close vicinity of the bait, but that may not be direct binders. Hence, in the present study, we investigated the STIM1 interactome using the BioID technique. A promiscuous biotin ligase was fused to the cytoplasmic C-terminus of STIM1 and was stably expressed in a mouse embryonic fibroblast (MEF) cell line. Screening of biotinylated proteins identified several high confidence targets. Here, we report Gelsolin (GSN) as a new member of the STIM1 interactome. GSN is a Ca2+-dependent actin-severing protein that promotes actin filament assembly and disassembly. Results were validated using knockdown approaches and immunostaining. We tested our results in neonatal cardiomyocytes where STIM1 overexpression induced altered actin dynamics and cytoskeletal instability. This is the first time that BioID assay was used to investigate the STIM1 interactome. Our work highlights the role of STIM1/GSN in the structure and function of the cytoskeleton.

Abstract 2156: Genomics and molecular analysis of Aurora kinase B (AURKB) in response to antiretroviral (ARV) treatment in lung cancer

Abstract Aurora kinase B (AURKB) is essential for normal physiological processes. However, aberrant AURKB expression has been reported in various cancers, including lung cancer. Lung cancer is a leading non-AIDS defining cancer (NADCs) and the relationship between the use of antiretrovirals (ARVs) and lung cancer is poorly understood. The anti-proliferative effects of Efavirenz (EFV) and Lopinavir/ritonavir (LPV/r) have also been documented. This study aimed at determining the genomic and molecular roles of AURKB in lung cancer in response to ARV treatment. For this purpose, cellular (xCELLigence RTCA), molecular (RT-qPCR) and bioinformatics tools were used in normal fibroblasts (MRC-5) and adenocarcinoma (A549) lung cells. Real-Time Cell Analysis (RTCA) data confirmed the anti-proliferative effects of ARVs in a time-dependent manner. AURKB gene expression was elevated in A549 compared to MRC-5 in non-ARV treated. Interestingly, p53 and CASP3 gene expression was downregulated. Contrarily, ARV treatment resulted in the formation of a negative signaling loop due to AURKB being downregulated, while CASP3 and p53 were upregulated. KEGG analysis demonstrated AURKB, CASP3 and p53 as cell cycle related genes. STRING database searches showed a strong protein-protein interaction (PPI) between AURKB and p53, and CAP3 and p53, but a weak PPI between AURKB and CASP3. Gene Ontology (GO) analysis revealed a decline in serine/threonine kinase (AURKB) activity in cancer cells in response to ARV treatment. Although not yet confirmed, AURKB displays oncogene features, while its overexpression has been shown to deregulate p53. The association between the use of ARVs and lung carcinogenesis remains to be fully comprehended and targeting AURKB signaling may help shed light on this relationship. Citation Format: Rahaba Makgotso Marima, Rodney Hull, Zodwa Dlamini, Clement Penny. Genomics and molecular analysis of Aurora kinase B (AURKB) in response to antiretroviral (ARV) treatment in lung cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 2156.

Protein Direct Interactome: Identification of Protein Direct Interactome with Genetic Code Expansion and Search Engine OpenUaa (Adv. Biology 3/2021)

Genetically encoded chemical crosslinking (GECX) via bioreactive unnatural amino acid (Uaa) enables the capture of weak and transient protein-protein interactions in live cells for identification with mass spectrometry. In article number 2000308, Chao Liu, Shang-Tong Li, Kui Li, Lei Wang, Bing Yang, and co-workers develop a new software OpenUaa to analyze Uaa-crosslinked peptides, which uniquely considers all fragment ions for high sensitivity and identify direct protein interactome with the GECX-OpenUaa strategy.

A tandem near-infrared fluorescence complementation system with enhanced fluorescence for imaging protein–protein interactions in vivo

Bimolecular fluorescence complementation (BiFC) is an effective tool for visualizing protein–protein interactions (PPIs). However, a BiFC system with long wavelength and high fluorescence intensity is yet to be developed for in vivo imaging. In this study, we constructed a tandem near-infrared BiFC (tBiFC) system by splitting a near-infrared phytochrome, IFP2.0. This system allows the identification and visualization of PPIs in live cells and living mice. The photophysical properties of the complementary fluorescence of the tBiFC system were similar to those of its parent protein IFP2.0, but the intensity was twice that of a single-copy IFP2.0-based BiFC system. Compared with previously reported near infrared BiFC systems—iRFP-BiFC and IFP1.4-BiFC—the signal intensity of the tBiFC system increased by ~1.48- and ~400-fold for weak PPIs in living cells, and ~1.51- and ~8-fold for strong PPIs, respectively. When applied to imaging PPIs in live mice, the complementary fluorescence intensity of the tBiFC system was also significantly higher than that of the other near-infrared BiFC systems. The use of this bright phytochrome in a tandem arrangement constitutes a powerful tool for imaging PPIs in the near infrared region.

Using Genetic Code Expansion for Protein Biochemical Studies.

Protein identification has gone beyond simply using protein/peptide tags and labeling canonical amino acids. Genetic code expansion has allowed residue- or site-specific incorporation of non-canonical amino acids into proteins. By taking advantage of the unique properties of non-canonical amino acids, we can identify spatiotemporal-specific protein states within living cells. Insertion of more than one non-canonical amino acid allows for selective labeling that can aid in the identification of weak or transient protein–protein interactions. This review will discuss recent studies applying genetic code expansion for protein labeling and identifying protein–protein interactions and offer considerations for future work in expanding genetic code expansion methods.

Liquid network connectivity regulates the stability and composition of biomolecular condensates with many components

One of the key mechanisms used by cells to control the spatiotemporal organization of their many components is the formation and dissolution of biomolecular condensates through liquid-liquid phase separation (LLPS). Using a minimal coarse-grained model that allows us to simulate thousands of interacting multivalent proteins, we investigate the physical parameters dictating the stability and composition of multicomponent biomolecular condensates. We demonstrate that the molecular connectivity of the condensed-liquid network-i.e., the number of weak attractive protein-protein interactions per unit of volume-determines the stability (e.g., in temperature, pH, salt concentration) of multicomponent condensates, where stability is positively correlated with connectivity. While the connectivity of scaffolds (biomolecules essential for LLPS) dominates the phase landscape, introduction of clients (species recruited via scaffold-client interactions) fine-tunes it by transforming the scaffold-scaffold bond network. Whereas low-valency clients that compete for scaffold-scaffold binding sites decrease connectivity and stability, those that bind to alternate scaffold sites not required for LLPS or that have higher-than-scaffold valencies form additional scaffold-client-scaffold bridges increasing stability. Proteins that establish more connections (via increased valencies, promiscuous binding, and topologies that enable multivalent interactions) support the stability of and are enriched within multicomponent condensates. Importantly, proteins that increase the connectivity of multicomponent condensates have higher critical points as pure systems or, if pure LLPS is unfeasible, as binary scaffold-client mixtures. Hence, critical points of accessible systems (i.e., with just a few components) might serve as a unified thermodynamic parameter to predict the composition of multicomponent condensates.

Physiological Tau Interactome in Brain and Its Link to Tauopathies.

Alzheimer's disease (AD) and most of the other tauopathies are incurable neurodegenerative diseases with unpleasant symptoms and consequences. The common hallmark of all of these diseases is tau pathology, but its connection with disease progress has not been completely understood so far. Therefore, uncovering novel tau-interacting partners and pathology affected molecular pathways can reveal the causes of diseases as well as potential targets for the development of AD treatment. Despite the large number of known tau-interacting partners, a limited number of studies focused on in vivo tau interactions in disease or healthy conditions are available. Here, we applied an in vivo cross-linking approach, capable of capturing weak and transient protein-protein interactions, to a unique transgenic rat model of progressive tau pathology similar to human AD. We have identified 175 potential novel and known tau-interacting proteins by MALDI-TOF mass spectrometry. Several of the most promising candidates for possible drug development were selected for validation by coimmunoprecipitation and colocalization experiments in animal and cellular models. Three proteins, Baiap2, Gpr37l1, and Nptx1, were confirmed as novel tau-interacting partners, and on the basis of their known functions and implications in neurodegenerative or psychiatric disorders, we proposed their potential role in tau pathology.

Photo-affinity pulling down of low-affinity binding proteins mediated by post-translational modifications

Weak and transient protein-protein interactions (PPIs) mediated by the post-translational modifications (PTMs) play key roles in biological systems. However, technical challenges to investigate the PTM-mediated PPIs have impeded many research advances. In this work, we develop a photo-affinity pull-down assay method to pull-down low-affinity binding proteins, thus for the screen of PTM-mediated PPIs. In this method, the PTM-mediated non-covalent interactions can be converted to the covalent interactions by the photo-activated linkage, so as to freeze frame the low-affinity binding interactions. The fabricated photo-affinity magnetic beads (PAMBs) ensure high specificity and resolution to capture the interacted proteins. Besides, the introduction of PEG passivation layer on PAMB has significantly reduced the non-specific interaction as compared to the traditional pull-down assay. For proof-of-concept, by using this newly developed assay method, we have identified a set of proteins that can interact with a specific methylation site on Flap Endonuclease 1 (FEN1) protein. Less interfering proteins (decreased over 80%) and more proteins sub-classes are profiled as compared to the traditional biotin-avidin pull-down system. Therefore, this new pull-down method may provide a useful tool for the study of low-affinity PPIs, and contribute to the discovery of potential targets for renewed PTM-mediated interactions that is fundamentally needed in biomedical research.

E. coli Single-Stranded DNA Binding (SSB) Protein Undergoes Dynamic Liquid-Liquid Phase Separation Controlled via Protein-Protein and Protein-DNA Interactions

A number of eukaryotic proteins have recently been shown to form phase-separated liquid condensates (membrane-less organelles) playing key roles in normal cell physiology and stress tolerance as bioreactors, biomolecular filters, stress sensors or molecular reservoirs. However, it is still debated whether liquid-liquid phase separation (LLPS) is a fundamental process in bacteria as it is in eukaryotes. Here we report the striking observation that the E. coli SSB protein, a ubiquitous central player of practically all DNA metabolic processes, forms phase-separated condensates under physiological conditions. LLPS is generally driven by multivalent weak protein-protein or protein-nucleic acid interactions, mediated mostly by intrinsically disordered protein regions. We found that LLPS by the homotetrameric SSB protein is mediated via multifaceted inter-tetramer interactions involving all protein regions including the conserved ssDNA-binding domain, the intrinsically disordered linker (IDL) and the C-terminal peptide (CTP, a highly conserved protein-protein interaction motif binding to a variety of DNA metabolic proteins). SSB, ssDNA and SSB-interacting partners are highly concentrated within the phase-separated droplets, whereas LLPS is overall regulated by the stoichiometry of SSB and ssDNA. Our bioinformatics analysis indicates that the LLPS-forming propensity of the SSB IDL is broadly conserved across all major phylogenetic groups of Eubacteria. Together with recently observed dynamic spot-like subcellular localization patterns of SSB, our results suggest that bacterial cells store an abundant pool of SSB and SSB-interacting proteins in phase-separated condensates via a conserved mechanism. The discovered features enable rapid mobilization of SSB to ssDNA regions exposed upon DNA damage or metabolic processes to serve efficient repair, replication and recombination.

Biophysical prediction of protein-peptide interactions and signaling networks using machine learning.

In mammalian cells, much of signal transduction is mediated by weak protein-protein interactions between globular peptide-binding domains (PBDs) and unstructured peptidic motifs in partner proteins. The number and diversity of these PBDs (over 1,800 are known), low binding affinities, and sensitivity of binding properties to minor sequence variation represent a substantial challenge to experimental and computational analysis of PBD specificity and the networks PBDs create. Here we introduce a bespoke machine learning approach, hierarchical statistical mechanical modelling (HSM), capable of accurately predicting the affinities of PBD-peptide interactions across multiple protein families. By synthesizing biophysical priors within a modern machine learning framework, HSM outperforms existing computational methods and high-throughput experimental assays. HSM models are interpretable in familiar biophysical terms at three spatial scales: the energetics of protein-peptide binding, the multi-dentate organization of protein-protein interactions, and the global architecture of signaling networks.