Identification Of Protein-protein Interactions Research Articles

Evaluating GPT and BERT models for protein-protein interaction identification in biomedical text.

Detecting protein-protein interactions (PPIs) is crucial for understanding genetic mechanisms, disease pathogenesis, and drug design. As biomedical literature continues to grow rapidly, there is an increasing need for automated and accurate extraction of these interactions to facilitate scientific discovery. Pretrained language models, such as generative pretrained transformers and bidirectional encoder representations from transformers, have shown promising results in natural language processing tasks. We evaluated the performance of PPI identification using multiple transformer-based models across three manually curated gold-standard corpora: Learning Language in Logic with 164 interactions in 77 sentences, Human Protein Reference Database with 163 interactions in 145 sentences, and Interaction Extraction Performance Assessment with 335 interactions in 486 sentences. Models based on bidirectional encoder representations achieved the best overall performance, with BioBERT achieving the highest recall of 91.95% and F1 score of 86.84% on the Learning Language in Logic dataset. Despite not being explicitly trained for biomedical texts, GPT-4 showed commendable performance, comparable to the bidirectional encoder models. Specifically, GPT-4 achieved the highest precision of 88.37%, a recall of 85.14%, and an F1 score of 86.49% on the same dataset. These results suggest that GPT-4 can effectively detect protein interactions from text, offering valuable applications in mining biomedical literature. The source code and datasets used in this study are available at https://github.com/hurlab/PPI-GPT-BERT.

Integrating Dynamic Network Analysis with AI for Enhanced Epitope Prediction in PD-L1:Affibody Interactions.

Understanding binding epitopes involved in protein-protein interactions and accurately determining their structure are long-standing goals with broad applicability in industry and biomedicine. Although various experimental methods for binding epitope determination exist, these approaches are typically low throughput and cost-intensive. Computational methods have potential to accelerate epitope predictions; however, recently developed artificial intelligence (AI)-based methods frequently fail to predict epitopes of synthetic binding domains with few natural homologues. Here we have developed an integrated method employing generalized-correlation-based dynamic network analysis on multiple molecular dynamics (MD) trajectories, initiated from AlphaFold2Multimer structures, to unravel the structure and binding epitope of the therapeutic PD-L1:Affibody complex. Both AlphaFold2 and conventional molecular dynamics trajectory analysis were ineffective in distinguishing between two proposed binding models, parallel and perpendicular. However, our integrated approach, utilizing dynamic network analysis, demonstrated that the perpendicular mode was significantly more stable. These predictions were validated using a suite of experimental epitope mapping protocols, including cross-linking mass spectrometry and next-generation sequencing-based deep mutational scanning. Conversely, AlphaFold3 failed to predict a structure bound in the perpendicular pose, highlighting the necessity for exploratory research in the search for binding epitopes and challenging the notion that AI-generated protein structures can be accepted without scrutiny. Our research underscores the potential of employing dynamic network analysis to enhance AI-based structure predictions for more accurate identification of protein-protein interaction interfaces.

African Swine Fever Virus Protein-Protein Interaction Prediction.

The African swine fever virus (ASFV) is an often deadly disease in swine and poses a threat to swine livestock and swine producers. With its complex genome containing more than 150 coding regions, developing effective vaccines for this virus remains a challenge due to a lack of basic knowledge about viral protein function and protein-protein interactions between viral proteins and between viral and host proteins. In this work, we identified ASFV-ASFV protein-protein interactions (PPIs) using artificial intelligence-powered protein structure prediction tools. We benchmarked our PPI identification workflow on the Vaccinia virus, a widely studied nucleocytoplasmic large DNA virus, and found that it could identify gold-standard PPIs that have been validated in vitro in a genome-wide computational screening. We applied this workflow to more than 18,000 pairwise combinations of ASFV proteins and were able to identify seventeen novel PPIs, many of which have corroborating experimental or bioinformatic evidence for their protein-protein interactions, further validating their relevance. Two protein-protein interactions, I267L and I8L, I267L__I8L, and B175L and DP79L, B175L__DP79L, are novel PPIs involving viral proteins known to modulate host immune response.

Allosteric regulation of the tyrosine phosphatase PTP1B by a protein-protein interaction.

The rapid identification of protein-protein interactions has been significantly enabled by mass spectrometry (MS) proteomics-based methods, including affinity purification-MS, crosslinking-MS, and proximity-labeling proteomics. While these methods can reveal networks of interacting proteins, they cannot reveal how specific protein-protein interactions alter cell signaling or protein function. For instance, when two proteins interact, there can be emergent signaling processes driven purely by the individual activities of those proteins being co-localized. Alternatively, protein-protein interactions can allosterically regulate function, enhancing or suppressing activity in response to binding. In this work, we investigate the interaction between the tyrosine phosphatase PTP1B and the adaptor protein Grb2, which have been annotated as binding partners in a number of proteomics studies. This interaction has been postulated to co-localize PTP1B with its substrate IRS-1 by forming a ternary complex, thereby enhancing the dephosphorylation of IRS-1 to suppress insulin signaling. Here, we report that Grb2 binding to PTP1B also allosterically enhances PTP1B catalytic activity. We show that this interaction is dependent on the proline-rich region of PTP1B, which interacts with the C-terminal SH3 domain of Grb2. Using NMR spectroscopy and hydrogen-deuterium exchange mass spectrometry (HDX-MS) we show that Grb2 binding alters PTP1B structure and/or dynamics. Finally, we use MS proteomics to identify other interactors of the PTP1B proline-rich region that may also regulate PTP1B function similarly to Grb2. This work presents one of the first examples of a protein allosterically regulating the enzymatic activity of PTP1B and lays the foundation for discovering new mechanisms of PTP1B regulation in cell signaling.

GNNGL-PPI: multi-category prediction of protein-protein interactions using graph neural networks based on global graphs and local subgraphs

Most proteins exert their functions by interacting with other proteins, making the identification of protein-protein interactions (PPI) crucial for understanding biological activities, pathological mechanisms, and clinical therapies. Developing effective and reliable computational methods for predicting PPI can significantly reduce the time-consuming and labor-intensive associated traditional biological experiments. However, accurately identifying the specific categories of protein-protein interactions and improving the prediction accuracy of the computational methods remain dual challenges. To tackle these challenges, we proposed a novel graph neural network method called GNNGL-PPI for multi-category prediction of PPI based on global graphs and local subgraphs. GNNGL-PPI consisted of two main components: using Graph Isomorphism Network (GIN) to extract global graph features from PPI network graph, and employing GIN As Kernel (GIN-AK) to extract local subgraph features from the subgraphs of protein vertices. Additionally, considering the imbalanced distribution of samples in each category within the benchmark datasets, we introduced an Asymmetric Loss (ASL) function to further enhance the predictive performance of the method. Through evaluations on six benchmark test sets formed by three different dataset partitioning algorithms (Random, BFS, DFS), GNNGL-PPI outperformed the state-of-the-art multi-category prediction methods of PPI, as measured by the comprehensive performance evaluation metric F1-measure. Furthermore, interpretability analysis confirmed the effectiveness of GNNGL-PPI as a reliable multi-category prediction method for predicting protein-protein interactions.

An economical, high-throughput protein-protein interaction modulator drug screening technique based on surface-enhanced Raman scattering

Increasing identification of protein-protein interaction (PPI) targets creates opportunities for drug discovery. However, the current drug screening strategies are costly and reagent-consuming, which hinders the progress in PPI modulator drug discovery. The primary obstacle lies in the inability to achieve molecular fishing with existing technology. Herein, we present a novel high-throughput, homogeneous-phase screening assay for PPI modulator discovery called “SERScreen”, based on a magnetic field-amplified surface-enhanced Raman spectroscopy (SERS). Two high-affinity proteins are fixed respectively on the magnetic beads (MBs) and the SERS tags, and the cross-linking of two nanoprobes induced by PPI enables strong SERS signals. Candidate modulators interfere with the PPI according to higher affinity toward one protein, resulting in a significant reduction in SERS intensity above the MBs. We established a potential drug screening platform for PD-1/PD-L1, and demonstrated its feasiblility not only with known inhibitors (Durvalumab and BMS-202), but also with a small-molecule combinatorial library, successfully identifying two new candidate inhibitors via the molecular fishing of SERScreen. Furthermore, the anticancer mechanism of two candidates was discussed. As an ultrasensitive, low reagent consumption (2 µL sample solution), high-throughput screening technique in the PPI modulator discovery, SERScreen offers an effective solution for molecular fishing from complex samples and displays high compatibility with automatic measurement equipment.

Advances in Computational Methods for Protein–Protein Interaction Prediction

Protein–protein interactions (PPIs) are pivotal in various physiological processes inside biological entities. Accurate identification of PPIs holds paramount significance for comprehending biological processes, deciphering disease mechanisms, and advancing medical research. Given the costly and labor-intensive nature of experimental approaches, a multitude of computational methods have been devised to enable swift and large-scale PPI prediction. This review offers a thorough examination of recent strides in computational methodologies for PPI prediction, with a particular focus on the utilization of deep learning techniques within this domain. Alongside a systematic classification and discussion of relevant databases, feature extraction strategies, and prominent computational approaches, we conclude with a thorough analysis of current challenges and prospects for the future of this field.

Identification of Protein-Protein Interaction Associated Functions Based on Gene Ontology.

Protein-protein interactions (PPIs) involve the physical or functional contact between two or more proteins. Generally, proteins that can interact with each other always have special relationships. Some previous studies have reported that gene ontology (GO) terms are related to the determination of PPIs, suggesting the special patterns on the GO terms of proteins in PPIs. In this study, we explored the special GO term patterns on human PPIs, trying to uncover the underlying functional mechanism of PPIs. The experimental validated human PPIs were retrieved from STRING database, which were termed as positive samples. Additionally, we randomly paired proteins occurring in positive samples, yielding lots of negative samples. A simple calculation was conducted to count the number of positive samples for each GO term pair, where proteins in samples were annotated by GO terms in the pair individually. The similar number for negative samples was also counted and further adjusted due to the great gap between the numbers of positive and negative samples. The difference of the above two numbers and the relative ratio compared with the number on positive samples were calculated. This ratio provided a precise evaluation of the occurrence of GO term pairs for positive samples and negative samples, indicating the latent GO term patterns for PPIs. Our analysis unveiled several nuclear biological processes, including gene transcription, cell proliferation, and nutrient metabolism, as key biological functions. Interactions between major proliferative or metabolic GO terms consistently correspond with significantly reported PPIs in recent literature.

Contribution of Artificial Intelligence to the Identification of Protein-Protein Interactions: A Case Study on PAR-3 and Its Partner Adapter Molecule Crk.

Protein-protein interactions (PPIs) are known to be involved in most cellular functions, and a detailed knowledge of such interactions is essential for studying their role in normal and pathological conditions. Significant progress is being made in the identification of PPIs through advances in computational methods. In particular, the AlphaFold2 machine learning-based model has been shown to accelerate drug discovery process by predicting the 3D structure of protein complexes. In this chapter, a straightforward protocol for predicting interprotein interactions between PAR-3 and its protein partner adapter molecule crk is provided. Such artificial intelligence-based and publicly available approaches can provide a resource for further investigation of therapeutic drug targets.

RGCNPPIS: A Residual Graph Convolutional Network for Protein-Protein Interaction Site Prediction.

Accurate identification of protein-protein interaction (PPI) sites is crucial for understanding the mechanisms of biological processes, developing PPI networks, and detecting protein functions. Currently, most computational methods primarily concentrate on sequence context features and rarely consider the spatial neighborhood features. To address this limitation, we propose a novel residual graph convolutional network for structure-based PPI site prediction (RGCNPPIS). Specifically, we use a GCN module to extract the global structural features from all spatial neighborhoods, and utilize the GraphSage module to extract local structural features from local spatial neighborhoods. To the best of our knowledge, this is the first work utilizing local structural features for PPI site prediction. We also propose an enhanced residual graph connection to combine the initial node representation, local structural features, and the previous GCN layer's node representation, which enables information transfer between layers and alleviates the over-smoothing problem. Evaluation results demonstrate that RGCNPPIS outperforms state-of-the-art methods on three independent test sets. In addition, the results of ablation experiments and case studies confirm that RGCNPPIS is an effective tool for PPI site prediction.

Analysis of Protein Interactions in Patient-Derived Xenografts Using Immunoprecipitation.

Proteins are large, complex molecules that regulate multiple functions within the cell. The protein rarely functions as a single molecule, but rather interacts with one or more other proteins forming a dynamic network. Protein-protein interactions are critical for regulating the cell's response toward various stimuli from outside and inside the cell. Identification of protein-protein interactions enhanced our understanding of various biological processes in the living cell. Immunoprecipitation (IP) has been one of the standard and most commonly used biochemical methods to identify and confirm protein-protein interactions. IP uses a target protein-specific antibody conjugated with protein A/G affinity beads to identify molecules interacting with the target protein. Here, we describe the principle, procedure and challenges of the IP assay.

Spatom: a graph neural network for structure-based protein-protein interaction site prediction.

Accurate identification of protein-protein interaction (PPI) sites remains a computational challenge. We propose Spatom, a novel framework for PPI site prediction. This framework first defines a weighted digraph for a protein structure to precisely characterize the spatial contacts of residues, then performs a weighted digraph convolution to aggregate both spatial local and global information and finally adds an improved graph attention layer to drive the predicted sites to form more continuous region(s). Spatom was tested on a diverse set of challenging protein-protein complexes and demonstrated the best performance among all the compared methods. Furthermore, when tested on multiple popular proteins in a case study, Spatom clearly identifies the interaction interfaces and captures the majority of hotspots. Spatom is expected to contribute to the understanding of protein interactions and drug designs targeting protein binding.

MS Annika 2.0 Identifies Cross-Linked Peptides in MS2-MS3-Based Workflows at High Sensitivity and Specificity.

Cross-linking mass spectrometry has become a powerful tool for the identification of protein-protein interactions and for gaining insight into the structures of proteins. We previously published MS Annika, a cross-linking search engine which can accurately identify cross-linked peptides in MS2 spectra from a variety of different MS-cleavable cross-linkers. In this publication, we present MS Annika 2.0, an updated version implementing a new search algorithm that, in addition to MS2 level, only supports the processing of data from MS2-MS3-based approaches for the identification of peptides from MS3 spectra, and introduces a novel scoring function for peptides identified across multiple MS stages. Detected cross-links are validated by estimating the false discovery rate (FDR) using a target-decoy approach. We evaluated the MS3-search-capabilities of MS Annika 2.0 on five different datasets covering a variety of experimental approaches and compared it to XlinkX and MaXLinker, two other cross-linking search engines. We show that MS Annika detects up to 4 times more true unique cross-links while simultaneously yielding less false positive hits and therefore a more accurate FDR estimation than the other two search engines. All mass spectrometry proteomics data along with result files have been deposited to the ProteomeXchange consortium via the PRIDE partner repository with the dataset identifier PXD041955.

High-throughput library transgenesis in Caenorhabditis elegans via Transgenic Arrays Resulting in Diversity of Integrated Sequences (TARDIS).

High-throughput transgenesis using synthetic DNA libraries is a powerful method for systematically exploring genetic function. Diverse synthesized libraries have been used for protein engineering, identification of protein-protein interactions, characterization of promoter libraries, developmental and evolutionary lineage tracking, and various other exploratory assays. However, the need for library transgenesis has effectively restricted these approaches to single-cell models. Here, we present Transgenic Arrays Resulting in Diversity of Integrated Sequences (TARDIS), a simple yet powerful approach to large-scale transgenesis that overcomes typical limitations encountered in multicellular systems. TARDIS splits the transgenesis process into a two-step process: creation of individuals carrying experimentally introduced sequence libraries, followed by inducible extraction and integration of individual sequences/library components from the larger library cassette into engineered genomic sites. Thus, transformation of a single individual, followed by lineage expansion and functional transgenesis, gives rise to thousands of genetically unique transgenic individuals. We demonstrate the power of this system using engineered, split selectable TARDIS sites in Caenorhabditis elegans to generate (1) a large set of individually barcoded lineages and (2) transcriptional reporter lines from predefined promoter libraries. We find that this approach increases transformation yields up to approximately 1000-fold over current single-step methods. While we demonstrate the utility of TARDIS using C. elegans, in principle the process is adaptable to any system where experimentally generated genomic loci landing pads and diverse, heritable DNA elements can be generated.

Chemical and Biological Strategies for Profiling Protein-Protein Interactions in Living Cells.

Protein-protein interactions (PPIs) play critical roles in almost all cellular signal transduction events. Characterization of PPIs without interfering with the functions of intact cells is very important for basic biology study and drug developments. However, the ability to profile PPIs especially those weak/transient interactions in their native states remains quite challenging. To this end, many endeavors are being made in developing new methods with high efficiency and strong operability. By coupling with advanced fluorescent microscopy and mass spectroscopy techniques, these strategies not only allow us to visualize the subcellular locations and monitor the functions of protein of interest (POI) in real time, but also enable the profiling and identification of potential unknown interacting partners in high-throughput manner, which greatly facilitates the elucidation of molecular mechanisms underlying numerous pathophysiological processes. In this review, we will summarize the typical methods for PPIs identification in living cells and their principles, advantages and limitations will also be discussed in detail.

Graph-BERT and language model-based framework for protein–protein interaction identification

Identification of protein–protein interactions (PPI) is among the critical problems in the domain of bioinformatics. Previous studies have utilized different AI-based models for PPI classification with advances in artificial intelligence (AI) techniques. The input to these models is the features extracted from different sources of protein information, mainly sequence-derived features. In this work, we present an AI-based PPI identification model utilizing a PPI network and protein sequences. The PPI network is represented as a graph where each node is a protein pair, and an edge is defined between two nodes if there exists a common protein between these nodes. Each node in a graph has a feature vector. In this work, we have used the language model to extract feature vectors directly from protein sequences. The feature vectors for protein in pairs are concatenated and used as a node feature vector of a PPI network graph. Finally, we have used the Graph-BERT model to encode the PPI network graph with sequence-based features and learn the hidden representation of the feature vector for each node. The next step involves feeding the learned representations of nodes to the fully connected layer, the output of which is fed into the softmax layer to classify the protein interactions. To assess the efficacy of the proposed PPI model, we have performed experiments on several PPI datasets. The experimental results demonstrate that the proposed approach surpasses the existing PPI works and designed baselines in classifying PPI.

Abstract 2021: Naveni TriFlex cell: An emerging method for simultaneous detection of free proteins and their interactions

Abstract Determining the levels of protein-protein interactions (PPI) is essential for the understanding of signal transduction, regulation of gene expression and mutation effects, infection mechanisms, identification of diagnostic markers, etc. Taking the already challenging task of PPI identification a step further, the next aim is to determine the levels of non-interacting proteins in parallel. The ability to concurrently monitor free and interacting proteins opens the door to studying their functional states and interplay, thus gaining deeper insight from a single staining experiment. Additionally, to accomplish this in situ, retaining the structural integrity of the cell, helps understand spatiotemporal communication between proteins in their native environment. To this end, we designed Naveni TriFlex Cell – a highly sensitive and specific proximity-based technology, relying on two user-determined primary antibodies against the targets of interest, and on proprietary TriFlex Navenibodies, which are antibody-based proximity reagents. TriFlex Cell detects total protein A (i.e., both free and in complex with B), total protein B, and the AB interaction. The detected A, B and AB signals are amplified and generate fluorescent readout in three channels corresponding to each protein pool. As proof of principle, we stained MCF7 cells for E-cadherin and β-catenin, which are known to interact in the adherens junctions. We observed highly abundant interactions at the cell membrane, some complexes and free proteins in the cytoplasm, and very low background in controls where either or both primary antibodies were omitted. These results attested to the high sensitivity and specificity of our method. Next, we explored the expression and interaction of Histone H3 and Lamin B, a nuclear envelope protein. Despite the compact structure of the nucleus, we were able to distinguish individual signals and abundant interactions, with free Lamin B particularly enriched in the nuclear membrane. In dividing cells, we observed diffuse Lamin signals in line with nuclear envelope breakdown, and more densely packed Histone H3 associated with chromatin. In contrast to non-mitotic cells, interactions were few, but increasing with the advance from metaphase to telophase. Finally, we stained against GM130, a Golgi complex protein, and COX1, a mitochondrial marker, as a biological non-interaction control experiment. TriFlex Cell successfully detected the individual proteins in their respective subcellular locations, and no complexes. In summary, our data demonstrate that Naveni TriFlex Cell is a sensitive and specific method that visualizes low and high abundant targets in various cell compartments and adds value by uncovering protein interplay, such as complex formation/dissolution under dynamic biological conditions. Citation Format: Axel Klaesson, Doroteya Raykova, Agata Zieba Wicher. Naveni TriFlex cell: An emerging method for simultaneous detection of free proteins and their interactions [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 2021.

Improving protein-protein interaction site prediction using deep residual neural network

Accurate identification of protein-protein interaction (PPI) sites is significantly important for understanding the mechanism of life and developing new drugs. However, it is expensive and time-consuming to identify PPI sites using wet-lab experiments. Developing computational methods is a new road to identify PPI sites, which can accelerate the procedure of PPI-related research. In this study, we propose a novel deep learning-based method (called D-PPIsite) to improve the accuracy of sequence-based PPI site prediction. In D-PPIsite, four discriminative sequence-driven features, i.e., position specific scoring matrix, relative solvent accessibility, position information and physical properties, are employed to feed into a well-designed deep learning module, consisting of convolutional, squeeze and excitation, and fully connected layers, to learn a prediction model. To reduce the risk of a single prediction model getting stuck in local optima, multiple prediction models with different initialization parameters are selected and integrated into one final model using the mean ensemble strategy. Experimental results on five independent testing data sets demonstrate that the proposed D-PPIsite can achieve an average accuracy of 80.2% and precision of 36.9%, covering 53.5% of all PPI sites while achieving the average Matthews correlation coefficient value (0.330) that is significantly higher than most of existing state-of-the-art prediction methods. We implement a new standalone-version predictor for predicting PPI sites, which is freely available at https://github.com/MingDongup/D-PPIsite for academic use.

Computational models for prediction of protein-protein interaction in rice and Magnaporthe grisea.

Plant-microbe interactions play a vital role in the development of strategies to manage pathogen-induced destructive diseases that cause enormous crop losses every year. Rice blast is one of the severe diseases to rice Oryza sativa (O. sativa) due to Magnaporthe grisea (M. grisea) fungus. Protein-protein interaction (PPI) between rice and fungus plays a key role in causing rice blast disease. In this paper, four genomic information-based models such as (i) the interolog, (ii) the domain, (iii) the gene ontology, and (iv) the phylogenetic-based model are developed for predicting the interaction between O. sativa and M. grisea in a whole-genome scale. A total of 59,430 interacting pairs between 1,801 rice proteins and 135 blast fungus proteins are obtained from the four models. Furthermore, a machine learning model is developed to assess the predicted interactions. Using composition-based amino acid composition (AAC) and conjoint triad (CT) features, an accuracy of 88% and 89% is achieved, respectively. When tested on the experimental dataset, the CT feature provides the highest accuracy of 95%. Furthermore, the specificity of the model is verified with other pathogen-host datasets where less accuracy is obtained, which confirmed that the model is specific to O. sativa and M. grisea. Understanding the molecular processes behind rice resistance to blast fungus begins with the identification of PPIs, and these predicted PPIs will be useful for drug design in the plant science community.

Dynamic Proximity Tagging in Living Plant Cells with Pupylation-Based Interaction Tagging.

Identification of protein-protein interactions (PPIs) and protein kinase substrates is fundamental for understanding how proteins exert biological functions with their partners and targets. However, it is still technically challenging, especially for transient and weak interactions involved in most cellular processes. The proximity-tagging systems enable capturing snapshots of both stable and transient PPIs. In this chapter, we describe in detail the methodology of a novel proximity-based labeling approach, PUP-IT (pupylation-based interaction tagging), to identify PPIs using a protoplast transient expression system. We have successfully identified potential kinase substrates by targeted screening and tandem mass spectrometry analysis.