Prediction Of Protein-protein Interactions Research Articles

Accelerating Protein-Protein interaction screens with reduced AlphaFold-Multimer sampling

Abstract Motivation Discovering new protein-protein interactions (PPIs) across entire proteomes offers vast potential for understanding novel protein functions and elucidate system properties within or between an organism. While recent advances in computational structural biology, particularly AlphaFold-Multimer, have facilitated this task, scaling for large-scale screenings remains a challenge, requiring significant computational resources. Results We evaluated the impact of reducing the number of models generated by AlphaFold-Multimer from five to one on the method's ability to distinguish true PPIs from false ones. Our evaluation was conducted on a dataset containing both intra- and inter-species PPIs, which included proteins from bacterial and eukaryotic sources. We demonstrate that reducing the sampling does not compromise the accuracy of the method, offering a faster, efficient, and environmentally friendly solution for PPI predictions. Availability and implementation The code used in this article is available at https://github.com/MIDIfactory/AlphaFastPPi. Note that the same can be achieved using the latest version of AlphaPulldown available at https://github.com/KosinskiLab/AlphaPulldown

Cimicifuga heracleifolia Kom. attenuates ulcerative colitis through the PI3K/AKT/NF-κB signaling pathway

Ethnopharmacological relevanceCimicifuga heracleifolia Kom. (C. heracleifolia) has demonstrated efficacy in treating gastrointestinal disorders, including splenasthenic diarrhea. Ulcerative colitis (UC), a chronic inflammatory bowel disease, shares similarities with splenasthenic diarrhea. However, the pharmacological effects of C. heracleifolia on UC and the underlying mechanisms remain unexplored.Aim of the study: The present study investigates the therapeutic potential and mechanisms of C. heracleifolia in UC. MethodsInitially, network pharmacology analysis, encompassing ingredient screening, target prediction, protein-protein interaction (PPI) network analysis, and enrichment analysis, was employed to predict the mechanisms of C. heracleifolia. The findings were further validated using transcriptomics and functional assays in a dextran sulfate sodium (DSS)-induced UC model. Additionally, bioactive compounds were identified through surface plasmon resonance (SPR) analysis, molecular docking, and cell-based assays. ResultsA total of 52 ingredients of C. heracleifolia were screened, and 32 key targets were identified within a PPI network comprising 285 potential therapeutic targets. Enrichment analysis indicated that the anti-UC effects of C. heracleifolia are mediated through immune response modulation and the inhibition of inflammatory signaling pathways. In vivo experiments showed that C. heracleifolia mitigated histological damage in the colon, reduced the expression of phosphorylated Akt1, nuclear factor-kappa B (NF-κB) p65, and inhibitor of Kappa B kinase α/β (IKKα/β), suppressed the content of interleukin-1β (IL-1β), interleukin-6 (IL-6), and tumor necrosis factor-α (TNF-α), and enhanced the expression of tight junction proteins. Moreover, cimigenoside, caffeic acid, and methyl caffeate were identified as the bioactive constituents responsible for the UC treatment effects of C. heracleifolia. ConclusionsIn summary, this study is the first to demonstrate that C. heracleifolia exerts therapeutic effects on UC by enhancing the intestinal mucosal barrier and inhibiting the phosphatidylinositol 3-kinase (PI3K)/AKT/NF-κB signaling pathway. These findings offer valuable insights into the clinical application of C. heracleifolia for UC management.

Combining handcrafted and learned features using deep learning to improve protein-protein interaction prediction performance

ABSTRACT Understanding protein-protein interactions (PPIs) is essential because knowledge regarding PPIs helps in determining the biochemical functions of organisms. Advances made in machine learning techniques, for example, DeepFE-PPI, GcForest-PPI, and DeepPPI, have been applied to enhance PPI prediction performance. However, most methods use either handcrafted or learned features. Improving protein representation quality can significantly impact PPI model performance. This study aims to propose a deep neural network-based model, in which both handcrafted and learned features are fused to improve PPI prediction. In our method, the handcrafted features are computed by sequence descriptors, including Amino Acid Composition (AAC), Local Descriptors, Pseudo-AAC, Amphiphilic Pseudo-AAC, and Quasi-Sequence-Order. The learned features are generated by Word2vec after being trained on a large database of protein sequences. Standard evaluation methods and benchmark datasets are used to assess the proposed model. The results show that our model achieves 95.8% and 99.3% accuracy on the Saccharomyces cerevisiae and Human datasets, respectively. Additionally, the generalizability of our model is evaluated through independent tests, and our findings indicate that the proposed feature combination method correctly classifies experimentally validated interactions in eight datasets, including cross-species and PPI networks. The source code and data for this work are included at https://github.com/npxquynhdhsp/ComFeatPPI.git.

Pair-EGRET: Enhancing the prediction of protein-protein interaction sites through graph attention networks and protein language models.

Proteins are responsible for most biological functions, many of which require the interaction of more than one protein molecule. However, accurately predicting protein-protein interaction (PPI) sites (the interfacial residues of a protein that interact with other protein molecules) remains a challenge. The growing demand and cost associated with the reliable identification of PPI sites using conventional experimental methods call for computational tools for automated prediction and understanding of PPIs. We present Pair-EGRET, an edge-aggregated graph attention network that leverages the features extracted from pre-trained transformer-like models to accurately predict PPI sites. Pair-EGRET works on a k-nearest neighbor graph, representing the three-dimensional structure of a protein, and utilizes the cross-attention mechanism for accurate identification of interfacial residues of a pair of proteins. Through an extensive evaluation study using a diverse array of experimental data, evaluation metrics, and case studies on representative protein sequences, we demonstrate that Pair-EGRET can achieve remarkable performance in predicting PPI sites. Moreover, Pair-EGRET can provide interpretable insights from the learned cross-attention matrix. Pair-EGRET is freely available in open source form at the GitHub Repository https://github.com/1705004/Pair-EGRET. Supplementary data are available at Bioinformatics online.

Identification of Homeobox Transcription Factors in a Dimorphic Fungus Talaromyces marneffei and Protein-Protein Interaction Prediction of RfeB

Talaromyces marneffei is a thermally dimorphic fungus that can cause life-threatening systemic mycoses, particularly in immunocompromised individuals. Fungal homeobox transcription factors control various developmental processes, including the regulation of sexual reproduction, morphology, metabolism, and virulence. However, the function of homeobox proteins in T. marneffei has not been fully explored. Here, we searched the T. marneffei genome for the total homeobox transcription factors and predicted their biological relevance by performing gene expression analysis in different cell types, including conidia, mycelia, yeasts, and during phase transition. RfeB is selected for further computational analysis since (i) its transcripts were differentially expressed in different phases of T. marneffei, and (ii) this protein contains the highly conserved protein-protein interaction region (IR), which could be important for pathobiology and have therapeutic application. To assess the structure-function of the IR region, in silico alanine substitutions were performed at three-conserved IR residues (Asp276, Glu279, and Gln282) of RfeB, generating a triple RfeB mutated protein. Using 3D modeling and molecular dynamics simulations, we compared the protein complex formation of wild-type and mutated RfeB proteins with the putative partner candidate TmSwi5. Our results demonstrated that the mutated RfeB protein exhibited increased free binding energy, elevated protein compactness, and a reduced number of atomic contacts, suggesting disrupted protein stability and interaction. Notably, our model revealed that the IR residues primarily stabilized the RfeB binding sites located in the central region (CR). This computational approach for protein mutagenesis could provide a foundation for future experimental studies on the functional characterization of RfeB and other homeodomain-containing proteins in T. marneffei.

Identifying nucleotide-binding leucine-rich repeat receptor and pathogen effector pairing using transfer-learning and bilinear attention network.

Nucleotide-binding leucine-rich repeat (NLR) family is a class of immune receptors capable of detecting and defending against pathogen invasion. They have been widely used in crop breeding. Notably, the correspondence between NLRs and effectors (CNE) determines the applicability and effectiveness of NLRs. Unfortunately, CNE data is very scarce. In fact, we've found a substantial 91,291 NLRs confirmed via wet experiments and bioinformatics methods but only 387 CNEs are recognized, which greatly restricts the potential application of NLRs. We propose a deep learning algorithm called ProNEP to identify NLR-effector pairs in a high-throughput manner. Specifically, we conceptualized the CNE prediction task as a protein-protein interaction (PPI) prediction task. Then, ProNEP predicts the interaction between NLRs and effectors by combining the transfer learning with a bilinear attention network. ProNEP achieves superior performance against state-of-the-art models designed for PPI predictions. Based on ProNEP, we conduct extensive identification of potential CNEs for 91,291 NLRs. With the rapid accumulation of genomic data, we expect that this tool will be widely used to predict CNEs in new species, advancing biology, immunology, and breeding. The ProNEP is available at http://nerrd.cn/#/prediction. The project code is available at https://github.com/ QiaoYJYJ/ProNEP. Supplementary data are available at Bioinformatics online.

Genome-wide identification, expression pattern and interacting protein analysis of INDETERMINATE DOMAIN (IDD) gene family in Phalaenopsis equestris

The plant-specific INDETERMINATE DOMAIN (IDD) gene family is important for plant growth and development. However, a comprehensive analysis of the IDD family in orchids is limited. Based on the genome data of Phalaenopsis equestris, the IDD gene family was identified and analyzed by bioinformatics methods in this study. Ten putative P. equestris IDD genes (PeIDDs) were characterized and phylogenetically classified into two groups according to their full amino acid sequences. Protein motifs analysis revealed that overall structures of PeIDDs in the same group were relatively conserved. Its promoter regions harbored a large number of responsive elements, including light responsive, abiotic stress responsive elements, and plant hormone cis-acting elements. The transcript level of PeIDD genes under cold and drought conditions, and by exogenous auxin (NAA) and abscisic acid (ABA) treatments further confirmed that most PeIDDs responded to various conditions and might play essential roles under abiotic stresses and hormone responses. In addition, distinct expression profiles in different tissues/organs suggested that PeIDDs might be involved in various development processes. Furthermore, the prediction of protein-protein interactions (PPIs) revealed some PeIDDs (PeIDD3 or PeIDD5) might function via cooperating with chromatin remodeling factors. The results of this study provided a reference for further understanding the function of PeIDDs.

Evolution of Apoptotic Signaling Pathways Within Lophotrochozoans.

Apoptosis is the main form of regulated cell death in metazoans. Apoptotic pathways are well characterized in nematodes, flies, and mammals, leading to a vision of the conservation of apoptotic pathways in metazoans. However, we recently showed that intrinsic apoptosis is in fact divergent among metazoans. In addition, extrinsic apoptosis is poorly studied in non-mammalian animals, making its evolution unclear. Consequently, our understanding of apoptotic signaling pathways evolution is a black box which must be illuminated by extending research to new biological systems. Lophotrochozoans are a major clade of metazoans which, despite their considerable biological diversity and key phylogenetic position as sister group of ecdysozoans (i.e. flies and nematodes), are poorly explored, especially regarding apoptosis mechanisms. Traditionally, each apoptotic signaling pathway was considered to rely on a specific initiator caspase, associated with an activator. To shed light on apoptosis evolution in animals, we explored the evolutionary history of initiator caspases, caspase activators, and the BCL-2 family (which control mitochondrial apoptotic pathway) in lophotrochozoans using phylogenetic analysis and protein interaction predictions. We discovered a diversification of initiator caspases in molluscs, annelids, and brachiopods, and the loss of key extrinsic apoptosis components in platyhelminths, along with the emergence of a clade-specific caspase with an ankyrin pro-domain. Taken together, our data show a specific history of apoptotic actors' evolution in lophotrochozoans, further demonstrating the appearance of distinct apoptotic signaling pathways during metazoan evolution.

Exploring the Effect of Fidgetin-Like 1 on Colorectal Cancer Through Tissue Chip and In Vitro Experiments.

Fidgetin-like 1 (FIGNL1) is extensively overexpressed in a variety of cancers. It facilitates non‑small cell lung cancer tumor cell proliferation and hepatocellular carcinoma formation due to abnormal DNA repair. Clinically relevant data indicates that its high expression is linked with the poor prognosis of patients with renal clear-cell carcinoma, low-grade gliomas, and hepatocellular carcinoma. Nevertheless, the scope of FIGNL1's involvement in cancer, particularly colorectal cancer (CRC), remains unclear. To investigate the function of FIGNL1 in CRC. The TCGA database and immunohistochemistry analysis were employed to investigate FIGNL1 expression in CRC tissue. A cell viability assay was performed using the Cell Counting Kit-8. The cell migration and invasion were evaluated using the transwell assay. Small interfering RNA (siRNA) transfection was conducted to knockdown FIGNL1 expression. Infection with FIGNL1 overexpression lentivirus was performed to promote FIGNL1 overexpression. The STRING database was employed for predicting protein interaction. FIGNL1 was substantially upregulated in human CRC tissues and was associated with TNM stages and lymph node metastasis in patients. The inhibition of CRC cell proliferation, migration, and invasion in Caco-2 cells was achieved by silencing FIGNL1 using siRNA. Additional investigations suggested that FIGNL1 overexpression could promote CRC cell proliferation, migration, and invasion via P38 signaling pathway activation in Colo-205 cells. Subsequent experiments demonstrated that FIGNL1-mediated P38 phosphorylation was contingent upon SPIDR interaction. These results implied that FIGNL1 was a potential anticancer drug target, which also offered a novel strategy for future CRC treatment.

Accurate and efficient protein embedding using multi-teacher distillation learning.

Protein embedding, which represents proteins as numerical vectors, is a crucial step in various learning-based protein annotation/classification problems, including gene ontology prediction, protein-protein interaction prediction, and protein structure prediction. However, existing protein embedding methods are often computationally expensive due to their large number of parameters, which can reach millions or even billions. The growing availability of large-scale protein datasets and the need for efficient analysis tools have created a pressing demand for efficient protein embedding methods. We propose a novel protein embedding approach based on multi-teacher distillation learning, which leverages the knowledge of multiple pre-trained protein embedding models to learn a compact and informative representation of proteins. Our method achieves comparable performance to state-of-the-art methods while significantly reducing computational costs and resource requirements. Specifically, our approach reduces computational time by ˜70% and maintains ±1.5% accuracy as the original large models. This makes our method well-suited for large-scale protein analysis and enables the bioinformatics community to perform protein embedding tasks more efficiently. The source code of MTDP is available via https://github.com/KennthShang/MTDP.

Predicting protein interactions of the kinase Lck critical to Tcell modulation.

Protein-protein interactions (PPIs) play pivotal roles in directing Tcell fate. One key player is the non-receptor tyrosine protein kinase Lck that helps to transduce Tcell activation signals. Lck is mediated by other proteins via interactions that are inadequately understood. Here, we use the deep learning method AF2Complex to predict PPIs involving Lck, by screening it against ∼1,000 proteins implicated in immune responses, followed by extensive structural modeling for selected interactions. Remarkably, we describe how Lck may be specifically targeted by a palmitoyltransferase using a phosphotyrosine motif. We uncover "hotspot" interactions between Lck and the tyrosine phosphatase CD45, leading to a significant conformational shift of Lck for activation. Lastly, we present intriguing interactions between the phosphotyrosine-binding domain of Lck and the cytoplasmic tail of the immune checkpoint LAG3 and propose a molecular mechanism for its inhibitory role. Together, this multifaceted study provides valuable insights into Tcell regulation and signaling.

Low CD3 level is a risk factor for amyotrophic lateral sclerosis: a Mendelian randomization study

Amyotrophic lateral sclerosis (ALS) is a progressive and fatal disease characterized by neuronal degeneration of the spinal cord and brain and believed to be related to the immune system. In this study, our aim is to use Mendelian randomization (MR) to search for immune markers related to ALS. A total of 731 immune cell traits were included in this study. MR analysis was used to identify the causality between 731 immune cell traits (with 3,757 Europeans) and ALS (with 138,086 Europeans). Colocalization analysis was used to verify the found causality, protein–protein interaction prediction was used to look for the interacting proteins that are known to be involved in ALS. We found low expression levels of CD3 on central memory CD8+ T cell is risk factor for ALS (OR = 0.90, 95% CI: 0.86–0.95, P = 0.0000303). CD3 can interact with three ALS-related proteins: VCP, HLA-DRA and HLA-DRB5, which are associated with adaptive immune response. Our study reported for the first time that low-level CD3 is a risk factor for ALS and the possible mechanism, which could provide a potential strategy for ALS diagnosis and therapy.

Genome-Wide Identification and Expression Analysis of the Cyclic Nucleotide-Gated Channel Gene Family in Zoysia japonica under Salt Stress.

Salt stress severely inhibits plant growth. Understanding the mechanism of plant salt tolerance is highly important to improving plant salt tolerance. Previous studies have shown that nonselective cyclic nucleotide-gated ion channels (CNGCs) play an important role in plant salt tolerance. However, current research on CNGCs mainly focuses on CNGCs in glycophytic plants, and research on CNGCs in halophytes that exhibit special salt tolerance strategies is still scarce. This study used the halophilic plant Zoysia japonica, an excellent warm-season turfgrass, as the experimental material. Through bioinformatics analysis, 18 members of the CNGC family were identified in Zoysia japonica; they were designated ZjCNGC1 through ZjCNGC18 according to their scaffold-level chromosomal positions. ZjCNGCs are divided into four groups (I-IV), with the same groups having differentiated protein-conserved domains and gene structures. ZjCNGCs are unevenly distributed on 16 scaffold-level chromosomes. Compared with other species, the ZjCNGCs in Group III exhibit obvious gene expansion, mainly due to duplication of gene segments. The collinearity between ZjCNGCs, OsCNGCs, and SjCNGCs suggests that CNGCs are evolutionarily conserved among gramineous plants. However, the Group III ZjCNGCs are only partially collinear with OsCNGCs and SjCNGCs, implying that the expansion of Group III ZjCNGC genes may have been an independent event occurring in Zoysia japonica. Protein interaction prediction revealed that ZjCNGCs, calcium-dependent protein kinase, H+-ATPase, outwardly rectifying potassium channel protein, and polyubiquitin 3 interact with ZjCNGCs. Multiple stress response regulatory elements, including those involved in salt stress, are present on the ZjCNGC promoter. The qPCR results revealed differences in the expression patterns of ZjCNGCs in different parts of the plant. Under salt stress conditions, the expression of ZjCNGCs was significantly upregulated in roots and leaves, with ZjCNGC8 and ZjCNGC13 showing the greatest increase in expression in the roots. These results collectively suggest that ZjCNGCs play an important role in salt tolerance and that their expansion into Group III may be a special mechanism underlying the salt tolerance of Zoysia japonica.

Intelligence model on sequence-based prediction of PPI using AISSO deep concept with hyperparameter tuning process

Protein–protein interaction (PPI) prediction is vital for interpreting biological activities. Even though many diverse sorts of data and machine learning approaches have been employed in PPI prediction, performance still has to be enhanced. As a result, we adopted an Aquilla Influenced Shark Smell (AISSO)-based hybrid prediction technique to construct a sequence-dependent PPI prediction model. This model has two stages of operation: feature extraction and prediction. Along with sequence-based and Gene Ontology features, unique features were produced in the feature extraction stage utilizing the improved semantic similarity technique, which may deliver reliable findings. These collected characteristics were then sent to the prediction step, and hybrid neural networks, such as the Improved Recurrent Neural Network and Deep Belief Networks, were used to predict the PPI using modified score level fusion. These neural networks’ weight variables were adjusted utilizing a unique optimal methodology called Aquila Influenced Shark Smell (AISSO), and the outcomes showed that the developed model had attained an accuracy of around 88%, which is much better than the traditional methods; this model AISSO-based PPI prediction can provide precise and effective predictions.

ESMDNN-PPI: a new protein–protein interaction prediction model developed with protein language model of ESM2 and deep neural network

ESMDNN-PPI: a new protein–protein interaction prediction model developed with protein language model of ESM2 and deep neural network

Genome-wide identification of oxidosqualene cyclase genes regulating natural rubber in Taraxacum kok-saghyz.

Nine TkOSC genes have been identified by genome-wide screening. Among them, TkOSC4-6 might be more crucial for natural rubber biosynthesis in Taraxacum kok-saghyz roots. Taraxacum kok-saghyz Rodin (TKS) roots contain large amounts of natural rubber, inulin, and valuable metabolites. Oxidosqualene cyclase (OSC) is a key member for regulating natural rubber biosynthesis (NRB) via the triterpenoid biosynthesis pathway. To explore the functions of OSC on natural rubber producing in TKS, its gene family members were identified in TKS genome via genome-wide screening. Nine TkOSCs were identified, which were mainly distributed in the cytoplasm. Their family genes experienced a neutral selection during the evolution process. Overall sequence homology analysis OSC proteins revealed 80.23% similarity, indicating a highly degree of conservation. Pairwise comparisons showed a multiple sequence similarity ranging from 57% to 100%. Protein interaction prediction revealed that TkOSCs may interact with baruol synthase, sterol 1,4-demethylase, lupeol synthase and squalene epoxidase. Phylogenetic analysis showed that OSC family proteins belong to two branches. TkOSC promoter regions contain cis-acting elements related to plant growth, stress response, hormones response and light response. Protein accumulation analysis demonstrated that TkOSC4, TkOSC5 and TkOSC6 proteins had strong expression levels in the root, latex and plumular axis. Comparison of gene expression patterns showed TkOSC1, TkOSC4, TkOSC5, TkOSC6, TkOSC7, TkOSC8 and TkOSC9 might be important in regulating NRB. Combination of gene and protein results revealed TkOSC4-6 might be more crucial, and the data might contribute to a more profound understanding of the roles of OSCs for NRB in TKS roots.

Protein-Protein Interaction Prediction Model Based on ProtBert-BiGRU-Attention.

The physiological activities within cells are mainly regulated through protein-protein interactions (PPI). Therefore, studying protein interactions has become an essential part of researching protein function and mechanisms. Traditional biological experiments required for PPI prediction are expensive and time consuming. For this reason, many methods based on predicting PPI from protein sequences have been proposed in recent years. However, existing computational methods usually require the combination of evolutionary feature information of proteins to predict PPI docking situations. Because different relevant features of selected proteins are chosen, there may be differences in the predicted results for PPI. This article proposes a PPI prediction method based on the pretrained protein sequence model ProtBert, combined with the Bidirectional Gated Recurrent Unit (BiGRU) and attention mechanism. Only using protein sequence information and leveraging ProtBert's powerful ability to capture amino acid feature information, BiGRU is used for further feature extraction of the amino acid vectors output by ProtBert. The attention mechanism is then applied to enhance the focus on different amino acid features and improve the expression ability of protein sequence features, ultimately obtaining binary classification results for protein interactions. Experimental results show that our proposed ProtBert-BiGRU-Attention model has good predictive performance for PPI. Through relevant comparative experiments, it has been proven that our model performs well in protein binary prediction. Furthermore, through the ablation experiment of the model, different deep learning modules' contributions to the prediction have been demonstrated.

Homeodomain leucine zipper (HD-ZIP) gene family in Apostasia shenzhenica Z. J. Liu & L. J. Chen (Orchidaceae): In silico identification and characterization

The homeodomain leucine zipper (HD-ZIP) gene family constitutes an important class of transcription factors found in plants that play crucial roles in various aspects of growth, development and stress regulation. Despite this, there have been minuscule reports on identification and characterization of HD-ZIP gene family in orchids. The present study identifies 31 putative genes (AsHD-ZIP) in Apostasia shenzhenica, using in silico approaches. All of the identified transcription factors were grouped into four subfamilies (HD-ZIP I to IV) based on the distribution of various conserved domains and motifs i.e., HD-ZIP I (homeodomain and leucine zipper), HD-ZIP II (homeodomain, leucine zipper and CPSCE), HD-ZIP III (homeodomain, leucine zipper, MEKHLA and START) and HD-ZIP IV (homeodomain, leucine zipper and START). Prediction of physicochemical properties, protein structures, protein-protein interactions, gene ontology, and gene structures further validated this classification. Phylogenetic analysis showed clustering of all subgroups separately along with their orthologs in Arabidopsis thaliana, Oryza sativa, Dendrobium catenatum and Phalaenopsis equestris. Furthermore, spatio-temporal expression profiling and presence of cis-regulatory elements confirmed tissue specific expression. As HD-ZIP genes are vital players in regulation of plant growth and development as well as in biotic and abiotic stress responses, this study provides a foundation for genetic improvement in orchids through functional validation and characterization of HD-ZIP gene family.

A comparison of antibody-antigen complex sequence-to-structure prediction methods and their systematic biases.

The ability to accurately predict antibody-antigen complex structures from their sequences could greatly advance our understanding of the immune system and would aid in the development of novel antibody therapeutics. There have been considerable recent advancements in predicting protein-protein interactions (PPIs) fueled by progress in machine learning (ML). To understand the current state of the field, we compare six representative methods for predicting antibody-antigen complexes from sequence, including two deep learning approaches trained to predict PPIs in general (AlphaFold-Multimer and RoseTTAFold), two composite methods that initially predict antibody and antigen structures separately and dock them (using antibody-mode ClusPro), local refinement in Rosetta (SnugDock) of globally docked poses from ClusPro, and a pipeline combining homology modeling with rigid-body docking informed by ML-based epitope and paratope prediction (AbAdapt). We find that AlphaFold-Multimer outperformed other methods, although the absolute performance leaves considerable room for improvement. AlphaFold-Multimer models of lower quality display significant structural biases at the level of tertiary motifs (TERMs) toward having fewer structural matches in non-antibody-containing structures from the Protein Data Bank (PDB). Specifically, better models exhibit more common PDB-like TERMs at the antibody-antigen interface than worse ones. Importantly, the clear relationship between performance and the commonness of interfacial TERMs suggests that the scarcity of interfacial geometry data in the structural database may currently limit the application of ML to the prediction of antibody-antigen interactions.

Predicting the Dynamic Interaction of Intrinsically Disordered Proteins.

Intrinsically disordered proteins (IDPs) participate in various biological processes. Interactions involving IDPs are usually dynamic and are affected by their inherent conformation fluctuations. Comprehensive characterization of these interactions based on current techniques is challenging. Here, we present GSALIDP, a GraphSAGE-embedded LSTM network, to capture the dynamic nature of IDP-involved interactions and predict their behaviors. This framework models multiple conformations of IDP as a dynamic graph, which can effectively describe the fluctuation of its flexible conformation. The dynamic interaction between IDPs is studied, and the data sets of IDP conformations and their interactions are obtained through atomistic molecular dynamic (MD) simulations. Residues of IDP are encoded through a series of features including their frustration. GSALIDP can effectively predict the interaction sites of IDP and the contact residue pairs between IDPs. Its performance in predicting IDP interactions is on par with or even better than the conventional models in predicting the interaction of structural proteins. To the best of our knowledge, this is the first model to extend the protein interaction prediction to IDP-involved interactions.