TIDGN: A Transfer Learning Framework for Predicting Interactions of Intrinsically Disordered Proteins with High Conformational Dynamics

Interactions between intrinsically disordered proteins (IDPs) are crucial for biological processes, such as intracellular liquid-liquid phase separation (LLPS). Experiments (e.g., NMR) and simulations used to study IDP interactions encounter a variety of difficulties, highlighting the necessity to develop relevant machine learning methods. However, reliable machine learning methods face the challenge resulting from the scarcity of available training data. In this work, we propose a transfer learning-based invariant geometric dynamic graph model, named TIDGN, for predicting IDP interactions. The model consists of a pretraining task module and a downstream task module. The pretraining task module learns the dynamic structural encoding of IDP monomers, which is then used by the downstream task module for interaction site prediction. The IDP monomer structure data set and the IDP interaction event data set are constructed using all-atom molecular dynamics (MD) simulations. The transfer learning strategy effectively enhances the model's performance. Both homotypic interactions and heterotypic interactions between two IDPs are considered in this work. Interestingly, TIDGN performs well for the heterotypic interaction prediction. Additionally, the feature ablation analysis emphasizes the importance of invariant geometric graph features. Taken together, our work demonstrates that the integration of transfer learning and the invariant geometric graph network offers a promising approach for addressing data scarcity challenges of IDP interaction prediction.

LGS-PPIS: A Local-Global Structural Information Aggregation Framework for Predicting Protein-Protein Interaction Sites

Exploring protein-protein interaction sites (PPIS) is of significance to elucidating the intrinsic mechanisms of diverse biological processes. On this basis, recent studies have applied deep learning-based technologies to overcome the high cost of wet experiments for PPIS determination. However, the existing methods still suffer from two limitations that remain to be solved. Firstly, the process of feature aggregation in most methods only took into account node features, but ignored the complex edge features of the target residue to its neighbor residues, resulting in insufficient local feature extraction. Secondly, such feature aggregation was limited to aggregating spatially adjacent residues, and could not capture the "remote" residues that played a critical role in determining PPIS, which can be summed up as the lack of global feature at the residue level. To break the above limitations, a local-global structural information aggregation framework, LGS-PPIS, was proposed in this study, including two modules of edge-aware graph convolutional network (EA-GCN) and self-attention integrated with initial residual and identity mapping (SA-RIM), which achieved the aggregation of local and global information for PPIS prediction. Evaluation results of LGS-PPIS showed that the proposed method outperformed state-of-the-art deep learning methods on three widely used PPIS prediction benchmarks. Besides, the results of ablation experiments demonstrated that the local features from spatially adjacent residues and global features from "remote" residues separately captured by EA-GCN and SA-RIM could benefit the model performance. Among them, the former was shown to have a more significant role in the PPIS prediction.

PIPENN-EMB ensemble net and protein embeddings generalise protein interface prediction beyond homology

Protein interactions are crucial for understanding biological functions and disease mechanisms, but predicting these remains a complex task in computational biology. Increasingly, Deep Learning models are having success in interface prediction. This study presents PIPENN-EMB which explores the added value of using embeddings from the ProtT5-XL protein language model. Our results show substantial improvement over the previously published PIPENN model for protein interaction interface prediction, reaching an MCC of 0.313 vs. 0.249, and AUROC 0.800 vs. 0.755 on the BIO_DL_TE test set. We furthermore show that these embeddings cover a broad range of 'hand-crafted' protein features in ablation studies. PIPENN-EMB reaches state-of-the-art performance on the ZK448 dataset for protein-protein interface prediction. We showcase predictions on 25 resistance-related proteins from Mycobacterium tuberculosis. Furthermore, whereas other state-of-the-art sequence-based methods perform worse for proteins that have little recognisable homology in their training data, PIPENN-EMB generalises to remote homologs, yielding stable AUROC across all three test sets with less than 30% sequence identity to the training dataset, and even to proteins with less than 15% sequence identity.

Two decades of advances in sequence-based prediction of MoRFs, disorder-to-order transitioning binding regions

INTRODUCTION

Molecular recognition features (MoRFs) are regions in protein sequences that undergo induced folding upon binding partner molecules. MoRFs are common in nature and can be predicted from sequences based on their distinctive sequence signatures.

AREAS COVERED

We overview 20 years of progress in the sequence-based prediction of MoRFs which resulted in the development of 25 predictors of MoRFs that interact with proteins, peptides, and lipids. These methods range from simple discriminant analysis to sophisticated deep transformer networks that use protein language models. They generate relatively accurate predictions as evidenced by the results of a recently published community-driven assessment.

EXPERT OPINION

MoRFs prediction is a mature field of research that is poised to continue at a steady pace in the foreseeable future. We anticipate further expansion of the scope of MoRF predictions to additional partner molecules, such as nucleic acids, and continued use of recent machine learning advances. Other future efforts should concentrate on improving availability of MoRF predictions by releasing, maintaining, and popularizing web servers and by depositing MoRF predictions to large databases of protein structure and function predictions. Furthermore, accurate MoRF predictions should be coupled with the equally accurate prediction and modeling of the resulting structures of complexes.

Accurate Prediction of Protein-Binding Residues in Protein Sequences Using SCRIBER

Deciphering molecular-level mechanisms that govern protein-protein interactions (PPIs) relies in part on the accurate prediction of protein-binding partners and protein-binding residues. These predictions can be used to support a wide spectrum of applications that include development of PPI networks and protein docking programs, drug design studies, and investigations of molecular details that underlie certain diseases. Computational methods that predict protein-binding residues offer convenient, inexpensive, and relatively accurate data that can aid these efforts. We introduce and describe a user-friendly webserver for the SCRIBER method that conveniently provides state-of-the-art predictions of protein-binding residues and that minimizes cross-predictions, i.e., incorrect prediction of residues that bind other/non-protein ligands as protein binding. SCRIBER relies on a two-layer architecture that is specifically designed to reduce the cross-predictions. We motivate and explain this predictive architecture. We describe how to use the webserver, interact with its web interface, and collect, read, and understand results generated by SCRIBER. The SCRIBER webserver is available at http://biomine.cs.vcu.edu/servers/SCRIBER/ .

DescribePROT Database of Residue-Level Protein Structure and Function Annotations

DescribePROT is a freely available online database of structural and functional descriptors of proteins at the amino acid level. It provides access to 13 diverse descriptors that include sequence conservation, putative secondary structure, solvent accessibility, intrinsic disorder, and signal peptides, and putative annotations of residues that interact with proteins, peptides and nucleic acids. These data can be used to elucidate protein functions, to support efforts to develop therapeutics, and to develop and evaluate future predictors of protein structure and function. DescribePROT includes 7.8 billion predictions for 1.4 million proteins from 83 complete proteomes of popular model organisms. This information can be downloaded at multiple levels of scope (entire database, specific organisms, and individual proteins) and can be interacted with using a graphical interface that simultaneously displays data on multiple descriptors. We describe the contents of this resource, provide directions on how to use its interface, and offer instructions on how to obtain and interact with the underlying data. Moreover, we briefly discuss plans for a future expansion of this database. DescribePROT is available at http://biomine.cs.vcu.edu/servers/DESCRIBEPROT/ .

Evaluation of predictions of disordered binding regions in the CAID2 experiment

A large portion of the Intrinsically Disordered Regions (IDRs) in protein sequences interact with proteins, nucleic acids, and other types of ligands. Correspondingly, dozens of sequence-based predictors of binding IDRs were developed. A recently completed second community-based Critical Assessments of protein Intrinsic Disorder prediction (CAID2) evaluated 32 predictors of binding IDRs. However, CAID2 considered a rather narrow scenario by testing on 78 proteins with binding IDRs and not differentiating between different ligands, in spite that virtually all predictors target IDRs that interact with specific types of ligands. In that scenario, several intrinsic disorder predictors predict binding IDRs with accuracy equivalent to the best predictors of binding IDRs since large majority of IDRs in the 78 test proteins are binding. We substantially extended the CAID2's evaluation by using the entire CAID2 dataset of 348 proteins and considering several arguably more practical scenarios. We assessed whether predictors accurately differentiate binding IDRs from other types of IDRs and how they perform when predicting IDRs that interact with different ligand types. We found that intrinsic disorder predictors cannot accurately identify binding IDRs among other disordered regions, majority of the predictors of binding IDRs are ligand type agnostic (i.e., they cross predict binding in IDRs that interact with ligands that they do not cover), and only a handful of predictors of binding IDRs perform relatively well and generate reasonably low amounts of cross predictions. We also suggest a number of future research directions that would move this active field of research forward.

GACT-PPIS: Prediction of protein-protein interaction sites based on graph structure and transformer network

The prediction of protein-protein interaction sites (PPIS) is currently crucial for regulating many biological activities in cells and developing drugs for various diseases. Deep learning-based methods have been proposed for predicting PPIS, significantly reducing the manpower and time costs associated with traditional experimental methods such as yeast two-hybrid, mass spectrometry, and affinity purification. However, the predictive accuracy of these deep learning methods has not yet reached the expected level. Therefore, we introduce a model called GACT-PPIS. The design of the GACT-PPIS algorithm aims to utilize combined information from protein sequences and structures as input to predict protein-protein interaction sites. The core of GACT-PPIS utilizes an Enhanced Graph Attention Network (EGAT) with initial residual and identity mappings, along with a deep Transformer network as the basic units, supplemented by Graph Convolutional Networks (GCN), effectively aggregating information from neighboring nodes for each node. After multiple network layers, the information of the entire protein is also fused into the nodes, and the Transformer network further enhances the model's performance. Experimental results show that GACT-PPIS outperforms the most representative models in terms of Recall, F1-measure, MCC, AUROC, and AUPRC on the benchmark test set (Test-60). Additionally, on other independent test sets (UBTest-31-6), GACT-PPIS leads in terms of Accuracy, Precision, Recall, F1-measure, MCC, AUROC, and AUPRC compared to the most representative models. It is worth noting that GACT-PPIS demonstrates excellent generalization and versatility across different test sets, showcasing good performance on multiple test sets for the same trained GACT-PPIS model.

Twenty years of advances in prediction of nucleic acid-binding residues in protein sequences

Computational prediction of nucleic acid-binding residues in protein sequences is an active field of research, with over 80 methods that were released in the past 2 decades. We identify and discuss 87 sequence-based predictors that include dozens of recently published methods that are surveyed for the first time. We overview historical progress and examine multiple practical issues that include availability and impact of predictors, key features of their predictive models, and important aspects related to their training and assessment. We observe that the past decade has brought increased use of deep neural networks and protein language models, which contributed to substantial gains in the predictive performance. We also highlight advancements in vital and challenging issues that include cross-predictions between deoxyribonucleic acid (DNA)-binding and ribonucleic acid (RNA)-binding residues and targeting the two distinct sources of binding annotations, structure-based versus intrinsic disorder-based. The methods trained on the structure-annotated interactions tend to perform poorly on the disorder-annotated binding and vice versa, with only a few methods that target and perform well across both annotation types. The cross-predictions are a significant problem, with some predictors of DNA-binding or RNA-binding residues indiscriminately predicting interactions with both nucleic acid types. Moreover, we show that methods with web servers are cited substantially more than tools without implementation or with no longer working implementations, motivating the development and long-term maintenance of the web servers. We close by discussing future research directions that aim to drive further progress in this area.

MERIT: Accurate Prediction of Multi Ligand-binding Residues with Hybrid Deep Transformer Network, Evolutionary Couplings and Transfer Learning

Multi-ligand binding residues (MLBRs) are amino acids in protein sequences that interact with multiple different ligands that include proteins, peptides, nucleic acids, and a variety of small molecules. MLBRs are implicated in a number of cellular functions and targeted in a context of multiple human diseases. There are many sequence-based predictors of residues that interact with specific ligand types and they can be collectively used to identify MLBRs. However, there are no methods that directly predict MLBRs. To this end, we conceptualize, design, evaluate and release MERIT (Multi-binding rEsidues pRedIcTor). This tool relies on a custom-crafted deep neural network that implements a number of innovative features, such as a multi-layered/step architecture with transformer modules that we train using a custom-designed loss function, computation of evolutionary couplings, and application of transfer learning. These innovations boost predictive performance, which we demonstrate using an ablation analysis. In particular, they reduce the number of cross-predictions, defined as residues that interact with a single ligand type that are incorrectly predicted as MLBRs. We compare MERIT against a representative selection of current and popular ligand-specific predictors, meta-predictors that combine their results to identify MLBRs, and a baseline regression-based predictor. These tests reveal that MERIT provides accurate predictions and statistically outperforms these alternatives. Moreover, using two test datasets, one with MLBRs and another with only the single ligand binding residues, we show that MERIT consistently produces relatively low false positive rates, including low rates of cross-predictions. The web server and datasets from this study are freely available at http://biomine.cs.vcu.edu/servers/MERIT/.

MIPPIS: protein-protein interaction site prediction network with multi-information fusion

BACKGROUND

The prediction of protein-protein interaction sites plays a crucial role in biochemical processes. Investigating the interaction between viruses and receptor proteins through biological techniques aids in understanding disease mechanisms and guides the development of corresponding drugs. While various methods have been proposed in the past, they often suffer from drawbacks such as long processing times, high costs, and low accuracy.

RESULTS

Addressing these challenges, we propose a novel protein-protein interaction site prediction network based on multi-information fusion. In our approach, the initial amino acid features are depicted by the position-specific scoring matrix, hidden Markov model, dictionary of protein secondary structure, and one-hot encoding. Simultaneously, we adopt a multi-channel approach to extract deep-level amino acids features from different perspectives. The graph convolutional network channel effectively extracts spatial structural information. The bidirectional long short-term memory channel treats the amino acid sequence as natural language, capturing the protein's primary structure information. The ProtT5 protein large language model channel outputs a more comprehensive amino acid embedding representation, providing a robust complement to the two aforementioned channels. Finally, the obtained amino acid features are fed into the prediction layer for the final prediction.

CONCLUSION

Compared with six protein structure-based methods and six protein sequence-based methods, our model achieves optimal performance across evaluation metrics, including accuracy, precision, F1, Matthews correlation coefficient, and area under the precision recall curve, which demonstrates the superiority of our model.

PMSFF: Improved Protein Binding Residues Prediction through Multi-Scale Sequence-Based Feature Fusion Strategy

Proteins perform different biological functions through binding with various molecules which are mediated by a few key residues and accurate prediction of such protein binding residues (PBRs) is crucial for understanding cellular processes and for designing new drugs. Many computational prediction approaches have been proposed to identify PBRs with sequence-based features. However, these approaches face two main challenges: (1) these methods only concatenate residue feature vectors with a simple sliding window strategy, and (2) it is challenging to find a uniform sliding window size suitable for learning embeddings across different types of PBRs. In this study, we propose one novel framework that could apply multiple types of PBRs Prediciton task through Multi-scale Sequence-based Feature Fusion (PMSFF) strategy. Firstly, PMSFF employs a pre-trained language model named ProtT5, to encode amino acid residues in protein sequences. Then, it generates multi-scale residue embeddings by applying multi-size windows to capture effective neighboring residues and multi-size kernels to learn information across different scales. Additionally, the proposed model treats protein sequences as sentences, employing a bidirectional GRU to learn global context. We also collect benchmark datasets encompassing various PBRs types and evaluate our PMSFF approach to these datasets. Compared with state-of-the-art methods, PMSFF demonstrates superior performance on most PBRs prediction tasks.

flDPnn2: Accurate and Fast Predictor of Intrinsic Disorder in Proteins

Prediction of the intrinsic disorder in protein sequences is an active research area, with well over 100 predictors that were released to date. These efforts are motivated by the functional importance and high levels of abundance of intrinsic disorder, combined with relatively low amounts of experimental annotations. The disorder predictors are periodically evaluated by independent assessors in the Critical Assessment of protein Intrinsic Disorder prediction (CAID) experiments. The recently completed CAID2 experiment assessed close to 40 state-of-the-art methods demonstrating that some of them produce accurate results. In particular, flDPnn2 method, which is the successor of flDPnn that performed well in the CAID1 experiment, secured the overall most accurate results on the Disorder-NOX dataset in CAID2. flDPnn2 implements a number of improvements when compared to its predecessor including changes to the inputs, increased size of the deep network model that we retrained on a larger training set, and addition of an alignment module. Using results from CAID2, we show that flDPnn2 produces accurate predictions very quickly, modestly improving over the accuracy of flDPnn and reducing the runtime by half, to about 27 s per protein. flDPnn2 is freely available as a convenient web server at http://biomine.cs.vcu.edu/servers/flDPnn2/.

Co-training based prediction of multi-label protein-protein interactions

Prediction of protein-protein interaction (PPI) types enhances the comprehension of the underlying structural characteristics and functions of proteins, which gives rise to a multi-label classification problem. The nominal features describe the physicochemical characteristics of proteins directly, establishing a more robust correlation with the interaction types between proteins than ordered features. Motivated by this, we propose a multi-label PPI prediction model referred to as CoMPPI (Co-training based Multi-Label prediction of Protein-Protein Interaction). This approach aims to maximize the utility of both ordered and nominal features extracted from protein sequences. Specifically, CoMPPI incorporates graph convolutional network (GCN) and 1D convolution operation to process the complementary subsets of features individually, leveraging both local and contextualized information in a more efficient way. In addition, two multi-type PPI datasets were constructed to eliminate the duplication in previous datasets. We compare the performance of CoMPPI with three state-of-the-art methods on three datasets partitioned using distinct schemes (Breadth-first search, Depth-first search, and Random), CoMPPI consistently outperforms the other methods across all cases, demonstrating improvements ranging from 3.81% to 32.40% in Micro-F1. The subsequent ablation experiment confirms the efficacy of employing the co-training framework for multi-label PPI prediction, indicating promising avenues for future advancements in this domain.

MR2CPPIS: Accurate prediction of protein-protein interaction sites based on multi-scale Res2Net with coordinate attention mechanism

Proteins play a vital role in various biological processes and achieve their functions through protein-protein interactions (PPIs). Thus, accurate identification of PPI sites is essential. Traditional biological methods for identifying PPIs are costly, labor-intensive, and time-consuming. The development of computational prediction methods for PPI sites offers promising alternatives. Most known deep learning (DL) methods employ layer-wise multi-scale CNNs to extract features from protein sequences. But, these methods usually neglect the spatial positions and hierarchical information embedded within protein sequences, which are actually crucial for PPI site prediction. In this paper, we propose MR2CPPIS, a novel sequence-based DL model that utilizes the multi-scale Res2Net with coordinate attention mechanism to exploit multi-scale features and enhance PPI site prediction capability. We leverage the multi-scale Res2Net to expand the receptive field for each network layer, thus capturing multi-scale information of protein sequences at a granular level. To further explore the local contextual features of each target residue, we employ a coordinate attention block to characterize the precise spatial position information, enabling the network to effectively extract long-range dependencies. We evaluate our MR2CPPIS on three public benchmark datasets (Dset 72, Dset 186, and PDBset 164), achieving state-of-the-art performance. The source codes are available at https://github.com/YyinGong/MR2CPPIS.

A comprehensive review of protein-centric predictors for biomolecular interactions: from proteins to nucleic acids and beyond

Proteins interact with diverse ligands to perform a large number of biological functions, such as gene expression and signal transduction. Accurate identification of these protein-ligand interactions is crucial to the understanding of molecular mechanisms and the development of new drugs. However, traditional biological experiments are time-consuming and expensive. With the development of high-throughput technologies, an increasing amount of protein data is available. In the past decades, many computational methods have been developed to predict protein-ligand interactions. Here, we review a comprehensive set of over 160 protein-ligand interaction predictors, which cover protein-protein, protein-nucleic acid, protein-peptide and protein-other ligands (nucleotide, heme, ion) interactions. We have carried out a comprehensive analysis of the above four types of predictors from several significant perspectives, including their inputs, feature profiles, models, availability, etc. The current methods primarily rely on protein sequences, especially utilizing evolutionary information. The significant improvement in predictions is attributed to deep learning methods. Additionally, sequence-based pretrained models and structure-based approaches are emerging as new trends.

AGF-PPIS: A protein-protein interaction site predictor based on an attention mechanism and graph convolutional networks

Protein-protein interactions play an important role in various biological processes. Interaction among proteins has a wide range of applications. Therefore, the correct identification of protein-protein interactions sites is crucial. In this paper, we propose a novel predictor for protein-protein interactions sites, AGF-PPIS, where we utilize a multi-head self-attention mechanism (introducing a graph structure), graph convolutional network, and feed-forward neural network. We use the Euclidean distance between each protein residue to generate the corresponding protein graph as the input of AGF-PPIS. On the independent test dataset Test_60, AGF-PPIS achieves superior performance over comparative methods in terms of seven different evaluation metrics (ACC, precision, recall, F1-score, MCC, AUROC, AUPRC), which fully demonstrates the validity and superiority of the proposed AGF-PPIS model. The source codes and the steps for usage of AGF-PPIS are available at https://github.com/fxh1001/AGF-PPIS.

HybridDBRpred: improved sequence-based prediction of DNA-binding amino acids using annotations from structured complexes and disordered proteins

Current predictors of DNA-binding residues (DBRs) from protein sequences belong to two distinct groups, those trained on binding annotations extracted from structured protein-DNA complexes (structure-trained) vs. intrinsically disordered proteins (disorder-trained). We complete the first empirical analysis of predictive performance across the structure- and disorder-annotated proteins for a representative collection of ten predictors. Majority of the structure-trained tools perform well on the structure-annotated proteins while doing relatively poorly on the disorder-annotated proteins, and vice versa. Several methods make accurate predictions for the structure-annotated proteins or the disorder-annotated proteins, but none performs highly accurately for both annotation types. Moreover, most predictors make excessive cross-predictions for the disorder-annotated proteins, where residues that interact with non-DNA ligand types are predicted as DBRs. Motivated by these results, we design, validate and deploy an innovative meta-model, hybridDBRpred, that uses deep transformer network to combine predictions generated by three best current predictors. HybridDBRpred provides accurate predictions and low levels of cross-predictions across the two annotation types, and is statistically more accurate than each of the ten tools and baseline meta-predictors that rely on averaging and logistic regression. We deploy hybridDBRpred as a convenient web server at http://biomine.cs.vcu.edu/servers/hybridDBRpred/ and provide the corresponding source code at https://github.com/jianzhang-xynu/hybridDBRpred.

Seq-InSite: sequence supersedes structure for protein interaction site prediction

MOTIVATION

Proteins accomplish cellular functions by interacting with each other, which makes the prediction of interaction sites a fundamental problem. As experimental methods are expensive and time consuming, computational prediction of the interaction sites has been studied extensively. Structure-based programs are the most accurate, while the sequence-based ones are much more widely applicable, as the sequences available outnumber the structures by two orders of magnitude. Ideally, we would like a tool that has the quality of the former and the applicability of the latter.

RESULTS

We provide here the first solution that achieves these two goals. Our new sequence-based program, Seq-InSite, greatly surpasses the performance of sequence-based models, matching the quality of state-of-the-art structure-based predictors, thus effectively superseding the need for models requiring structure. The predictive power of Seq-InSite is illustrated using an analysis of evolutionary conservation for four protein sequences.

AVAILABILITY AND IMPLEMENTATION

Seq-InSite is freely available as a web server at http://seq-insite.csd.uwo.ca/ and as free source code, including trained models and all datasets used for training and testing, at https://github.com/lucian-ilie/Seq-InSite.

GHGPR-PPIS: A graph convolutional network for identifying protein-protein interaction site using heat kernel with Generalized PageRank techniques and edge self-attention feature processing block

Accurately pinpointing protein-protein interaction site (PPIS) on the molecular level is of utmost significance for annotating protein function and comprehending the mechanisms underpinning various diseases. While numerous computational methods for predicting PPIS have emerged, they have indeed mitigated the labor and time constraints associated with traditional experimental methods. However, the predictive accuracy of these methods has yet to reach the desired threshold. In this context, we proposed a groundbreaking graph-based computational model called GHGPR-PPIS. This innovative model leveraged a graph convolutional network using heat kernel (GraphHeat) in conjunction with Generalized PageRank techniques (GHGPR) to predict PPIS. Additionally, building upon the GHGPR framework, we devised an edge self-attention feature processing block, further augmenting the performance of the model. Experimental findings conclusively demonstrated that GHGPR-PPIS surpassed all competing state-of-the-art models when evaluated on the benchmark test set. Impressively, on two distinct independent test sets and a specific protein chain, GHGPR-PPIS consistently demonstrated superior generalization performance and practical applicability compared to the comparative model, AGAT-PPIS. Lastly, leveraging the t-SNE dimensionality reduction algorithm and clustering visualization technique, we delved into an interpretability analysis of the effectiveness of GHGPR-PPIS by meticulously comparing the outputs from different stages of the model.

A Transformer-Based Ensemble Framework for the Prediction of Protein-Protein Interaction Sites

The identification of protein-protein interaction (PPI) sites is essential in the research of protein function and the discovery of new drugs. So far, a variety of computational tools based on machine learning have been developed to accelerate the identification of PPI sites. However, existing methods suffer from the low predictive accuracy or the limited scope of application. Specifically, some methods learned only global or local sequential features, leading to low predictive accuracy, while others achieved improved performance by extracting residue interactions from structures but were limited in their application scope for the serious dependence on precise structure information. There is an urgent need to develop a method that integrates comprehensive information to realize proteome-wide accurate profiling of PPI sites. Herein, a novel ensemble framework for PPI sites prediction, EnsemPPIS, was therefore proposed based on transformer and gated convolutional networks. EnsemPPIS can effectively capture not only global and local patterns but also residue interactions. Specifically, EnsemPPIS was unique in (a) extracting residue interactions from protein sequences with transformer and (b) further integrating global and local sequential features with the ensemble learning strategy. Compared with various existing methods, EnsemPPIS exhibited either superior performance or broader applicability on multiple PPI sites prediction tasks. Moreover, pattern analysis based on the interpretability of EnsemPPIS demonstrated that EnsemPPIS was fully capable of learning residue interactions within the local structure of PPI sites using only sequence information. The web server of EnsemPPIS is freely available at http://idrblab.org/ensemppis.

Accurately identifying nucleic-acid-binding sites through geometric graph learning on language model predicted structures

The interactions between nucleic acids and proteins are important in diverse biological processes. The high-quality prediction of nucleic-acid-binding sites continues to pose a significant challenge. Presently, the predictive efficacy of sequence-based methods is constrained by their exclusive consideration of sequence context information, whereas structure-based methods are unsuitable for proteins lacking known tertiary structures. Though protein structures predicted by AlphaFold2 could be used, the extensive computing requirement of AlphaFold2 hinders its use for genome-wide applications. Based on the recent breakthrough of ESMFold for fast prediction of protein structures, we have developed GLMSite, which accurately identifies DNA- and RNA-binding sites using geometric graph learning on ESMFold predicted structures. Here, the predicted protein structures are employed to construct protein structural graph with residues as nodes and spatially neighboring residue pairs for edges. The node representations are further enhanced through the pre-trained language model ProtTrans. The network was trained using a geometric vector perceptron, and the geometric embeddings were subsequently fed into a common network to acquire common binding characteristics. Finally, these characteristics were input into two fully connected layers to predict binding sites with DNA and RNA, respectively. Through comprehensive tests on DNA/RNA benchmark datasets, GLMSite was shown to surpass the latest sequence-based methods and be comparable with structure-based methods. Moreover, the prediction was shown useful for inferring nucleic-acid-binding proteins, demonstrating its potential for protein function discovery. The datasets, codes, and trained models are available at https://github.com/biomed-AI/nucleic-acid-binding.

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.

E(3) equivariant graph neural networks for robust and accurate protein-protein interaction site prediction

Artificial intelligence-powered protein structure prediction methods have led to a paradigm-shift in computational structural biology, yet contemporary approaches for predicting the interfacial residues (i.e., sites) of protein-protein interaction (PPI) still rely on experimental structures. Recent studies have demonstrated benefits of employing graph convolution for PPI site prediction, but ignore symmetries naturally occurring in 3-dimensional space and act only on experimental coordinates. Here we present EquiPPIS, an E(3) equivariant graph neural network approach for PPI site prediction. EquiPPIS employs symmetry-aware graph convolutions that transform equivariantly with translation, rotation, and reflection in 3D space, providing richer representations for molecular data compared to invariant convolutions. EquiPPIS substantially outperforms state-of-the-art approaches based on the same experimental input, and exhibits remarkable robustness by attaining better accuracy with predicted structural models from AlphaFold2 than what existing methods can achieve even with experimental structures. Freely available at https://github.com/Bhattacharya-Lab/EquiPPIS, EquiPPIS enables accurate PPI site prediction at scale.

Clinical interest of molecular study in cases of isolated midline craniosynostosis

In some cases of infants with apparently isolated single-suture synostosis, an underlying variant can be found. We aimed to determine the molecular substratum in isolated sagittal and metopic craniosynostosis. To this end, we included all infants who presented isolated midline synostosis (sagittal or metopic) and had undergone surgery at the craniosynostosis national reference center of Lyon University Hospital. All infants were examined by a multidisciplinary team including neurosurgeons, clinical geneticists and neuropsychologist. Among 101 infants tested, 13 carried a total of 13 variants; that is, 12.9% of the infants carried a variant in genes known to be involved in craniosynostosis. Seven infants carried SMAD6 variants, 2 in FGFR2, 1 in TWIST1, one in FREM1, one in ALX4 and one in TCF12. All variants were detected at the heterozygous level in genes associated with autosomal dominant craniosynostosis. Also, neurodevelopmental testing showed especially delayed acquisition of language in children with than without variants in SMAD6. In conclusion, a high percentage of young children with isolated midline craniosynostosis, especially in isolated trigonocephaly, carried SMAD6 variants. The interpretation of the pathogenicity of the genes must take into account incomplete penetrance, usually observed in craniosynostosis. Our results highlight the interest of molecular analysis in the context of isolated sagittal and/or metopic craniosynostosis to enhance an understanding of the pathophysiology of midline craniosynostosis.

LigBind: identifying binding residues for over 1000 ligands with relation-aware graph neural networks

Identifying the interactions between proteins and ligands is significant for drug discovery and design. Considering the diverse binding patterns of ligands, the ligand-specific methods are trained per ligand to predict binding residues. However, most of the existing ligand-specific methods ignore shared binding preferences among various ligands and generally only cover a limited number of ligands with a sufficient number of known binding proteins. In this study, we propose a relation-aware framework LigBind with graph-level pre-training to enhance the ligand-specific binding residue predictions for 1159 ligands, which can effectively cover the ligands with a few known binding proteins. LigBind first pre-trains a graph neural network-based feature extractor for ligand-residue pairs and relation-aware classifiers for similar ligands. Then, LigBind is fine-tuned with ligand-specific binding data, where a domain adaptive neural network is designed to automatically leverage the diversity and similarity of various ligand-binding patterns for accurate binding residue prediction. We construct ligand-specific benchmark datasets of 1159 ligands and 16 unseen ligands, which are used to evaluate the effectiveness of LigBind. The results demonstrate the LigBind's efficacy on the large-scale ligand-specific benchmark datasets, and generalizes well to unseen ligands. LigBind also enables accurate identification of the ligand-binding residues in the main protease, papain-like protease and the RNA-dependent RNA polymerase of SARS-CoV-2. The webserver and source codes of LigBind are available at http://www.csbio.sjtu.edu.cn/bioinf/LigBind/ and https://github.com/YYingXia/LigBind/ for academic use.

HybridRNAbind: prediction of RNA interacting residues across structure-annotated and disorder-annotated proteins

The sequence-based predictors of RNA-binding residues (RBRs) are trained on either structure-annotated or disorder-annotated binding regions. A recent study of predictors of protein-binding residues shows that they are plagued by high levels of cross-predictions (protein binding residues are predicted as nucleic acid binding) and that structure-trained predictors perform poorly for the disorder-annotated regions and vice versa. Consequently, we analyze a representative set of the structure and disorder trained predictors of RBRs to comprehensively assess quality of their predictions. Our empirical analysis that relies on a new and low-similarity benchmark dataset reveals that the structure-trained predictors of RBRs perform well for the structure-annotated proteins while the disorder-trained predictors provide accurate results for the disorder-annotated proteins. However, these methods work only modestly well on the opposite types of annotations, motivating the need for new solutions. Using an empirical approach, we design HybridRNAbind meta-model that generates accurate predictions and low amounts of cross-predictions when tested on data that combines structure and disorder-annotated RBRs. We release this meta-model as a convenient webserver which is available at https://www.csuligroup.com/hybridRNAbind/.

SCAMPER: Accurate Type-Specific Prediction of Calcium-Binding Residues Using Sequence-Derived Features

Understanding molecular mechanisms involved in calcium-protein interactions and modeling corresponding docking rely on the accurate identification of calcium-binding residues (CaBRs). The defects of experimentally annotating protein functions enhances the development of computational approaches that correctly identify calcium-binding interactions. Studies have reported that current methods severely cross-predict residues that interact with other types of molecules (e.g., nucleic acids, proteins, and small ligands) as CaBRs. In this study, a novel predictor named SCAMPER (Selective CAlciuM-binding PrEdictoR) is proposed for the accurate and specific prediction of CaBRs. SCAMPER is designed using newly compiled dataset with complete UniProt sequences and annotations, which include calcium-binding, nucleic acid-binding, protein-binding, and small ligand-binding residues. We use a novel designed two-layer scheme to perform predictions as well as penalize cross-predictions. Empirical tests on an independent test dataset reveals that the proposed method significantly outperforms state-of-the-art predictors. SCAMPER is proved to be capable of distinguishing CaBRs from different types of metal-ion binding residues. We further perform CaBRs predictions on the whole human proteome, and use the results to hypothesize calcium-binding proteins (CaBPs). The latest experimental verified CaBPs and GO analysis prove the accuracy of our predictions. We implement the proposed method and share the data at http://www.inforstation.com/webservers/SCAMPER/.

Computational prediction of disordered binding regions

One of the key features of intrinsically disordered regions (IDRs) is their ability to interact with a broad range of partner molecules. Multiple types of interacting IDRs were identified including molecular recognition fragments (MoRFs), short linear sequence motifs (SLiMs), and protein-, nucleic acids- and lipid-binding regions. Prediction of binding IDRs in protein sequences is gaining momentum in recent years. We survey 38 predictors of binding IDRs that target interactions with a diverse set of partners, such as peptides, proteins, RNA, DNA and lipids. We offer a historical perspective and highlight key events that fueled efforts to develop these methods. These tools rely on a diverse range of predictive architectures that include scoring functions, regular expressions, traditional and deep machine learning and meta-models. Recent efforts focus on the development of deep neural network-based architectures and extending coverage to RNA, DNA and lipid-binding IDRs. We analyze availability of these methods and show that providing implementations and webservers results in much higher rates of citations/use. We also make several recommendations to take advantage of modern deep network architectures, develop tools that bundle predictions of multiple and different types of binding IDRs, and work on algorithms that model structures of the resulting complexes.

Learning the protein language of proteome-wide protein-protein binding sites via explainable ensemble deep learning

Protein-protein interactions (PPIs) govern cellular pathways and processes, by significantly influencing the functional expression of proteins. Therefore, accurate identification of protein-protein interaction binding sites has become a key step in the functional analysis of proteins. However, since most computational methods are designed based on biological features, there are no available protein language models to directly encode amino acid sequences into distributed vector representations to model their characteristics for protein-protein binding events. Moreover, the number of experimentally detected protein interaction sites is much smaller than that of protein-protein interactions or protein sites in protein complexes, resulting in unbalanced data sets that leave room for improvement in their performance. To address these problems, we develop an ensemble deep learning model (EDLM)-based protein-protein interaction (PPI) site identification method (EDLMPPI). Evaluation results show that EDLMPPI outperforms state-of-the-art techniques including several PPI site prediction models on three widely-used benchmark datasets including Dset_448, Dset_72, and Dset_164, which demonstrated that EDLMPPI is superior to those PPI site prediction models by nearly 10% in terms of average precision. In addition, the biological and interpretable analyses provide new insights into protein binding site identification and characterization mechanisms from different perspectives. The EDLMPPI webserver is available at http://www.edlmppi.top:5002/ .

DeepPRObind: Modular Deep Learner that Accurately Predicts Structure and Disorder-Annotated Protein Binding Residues

Current sequence-based predictors of protein-binding residues (PBRs) belong to two distinct categories: structure-trained vs. intrinsic disorder-trained. Since disordered PBRs differ from structured PBRs in several ways, including ability to bind multiple partners by folding into different conformations and enrichment in different amino acids, the structure-trained and disorder-trained predictors were shown to provide inaccurate results for the other annotation type. A simple consensus-based solution that combines structure- and disorder-trained methods provides limited levels of predictive performance and generates relatively many cross-predictions, where residues that interact with other ligand types are predicted as PBRs. We address this unsolved problem by designing a novel and fast deep-learner, DeepPRObind, that relies on carefully designed modular convolutional architecture and uses innovative aggregate input features. Comparative empirical tests on a low-similarity test dataset reveal that DeepPRObind generates accurate predictions of structured and disordered PBRs and low amounts of cross-predictions, outperforming a comprehensive collection of 12 predictors of PBRs. Given the relatively low runtime of DeepPRObind (40 seconds per protein), we further validate its results based on an analysis of putative PBRs in the yeast proteome, confirming that interactions in disordered regions are enriched among hub proteins. We release DeepPRObind as a convenient web server at https://www.csuligroup.com/DeepPRObind/.

BindWeb: A web server for ligand binding residue and pocket prediction from protein structures

Knowledge of protein-ligand interactions is beneficial for biological process analysis and drug design. Given the complexity of the interactions and the inadequacy of experimental data, accurate ligand binding residue and pocket prediction remains challenging. In this study, we introduce an easy-to-use web server BindWeb for ligand-specific and ligand-general binding residue and pocket prediction from protein structures. BindWeb integrates a graph neural network GraphBind with a hybrid convolutional neural network and bidirectional long short-term memory network DELIA to identify binding residues. Furthermore, BindWeb clusters the predicted binding residues to binding pockets with mean shift clustering. The experimental results and case study demonstrate that BindWeb benefits from the complementarity of two base methods. BindWeb is freely available for academic use at http://www.csbio.sjtu.edu.cn/bioinf/BindWeb/.

Ten quick tips for sequence-based prediction of protein properties using machine learning

The ubiquitous availability of genome sequencing data explains the popularity of machine learning-based methods for the prediction of protein properties from their amino acid sequences. Over the years, while revising our own work, reading submitted manuscripts as well as published papers, we have noticed several recurring issues, which make some reported findings hard to understand and replicate. We suspect this may be due to biologists being unfamiliar with machine learning methodology, or conversely, machine learning experts may miss some of the knowledge needed to correctly apply their methods to proteins. Here, we aim to bridge this gap for developers of such methods. The most striking issues are linked to a lack of clarity: how were annotations of interest obtained; which benchmark metrics were used; how are positives and negatives defined. Others relate to a lack of rigor: If you sneak in structural information, your method is not sequence-based; if you compare your own model to "state-of-the-art," take the best methods; if you want to conclude that some method is better than another, obtain a significance estimate to support this claim. These, and other issues, we will cover in detail. These points may have seemed obvious to the authors during writing; however, they are not always clear-cut to the readers. We also expect many of these tips to hold for other machine learning-based applications in biology. Therefore, many computational biologists who develop methods in this particular subject will benefit from a concise overview of what to avoid and what to do instead.

Multi-task learning to leverage partially annotated data for PPI interface prediction

Protein protein interactions (PPI) are crucial for protein functioning, nevertheless predicting residues in PPI interfaces from the protein sequence remains a challenging problem. In addition, structure-based functional annotations, such as the PPI interface annotations, are scarce: only for about one-third of all protein structures residue-based PPI interface annotations are available. If we want to use a deep learning strategy, we have to overcome the problem of limited data availability. Here we use a multi-task learning strategy that can handle missing data. We start with the multi-task model architecture, and adapted it to carefully handle missing data in the cost function. As related learning tasks we include prediction of secondary structure, solvent accessibility, and buried residue. Our results show that the multi-task learning strategy significantly outperforms single task approaches. Moreover, only the multi-task strategy is able to effectively learn over a dataset extended with structural feature data, without additional PPI annotations. The multi-task setup becomes even more important, if the fraction of PPI annotations becomes very small: the multi-task learner trained on only one-eighth of the PPI annotations—with data extension—reaches the same performances as the single-task learner on all PPI annotations. Thus, we show that the multi-task learning strategy can be beneficial for a small training dataset where the protein’s functional properties of interest are only partially annotated.

Predicting protein-peptide binding residues via interpretable deep learning

 

Identifying the protein-peptide binding residues is fundamentally important to understand the mechanisms of protein functions and explore drug discovery. Although several computational methods have been developed, they highly rely on third-party tools or information for feature design, easily resulting in low computational efficacy and suffering from low predictive performance. To address the limitations, we propose PepBCL, a novel BERT (Bidirectional Encoder Representation from Transformers)-based Contrastive Learning framework to predict the protein-Peptide binding residues based on protein sequences only. PepBCL is an end-to-end predictive model that is independent of designed features. Specifically, we introduce a well pre-trained protein language model that can automatically extract and learn high-latent representations of protein sequences relevant for protein structure and functions. Further, we design a novel contrastive learning module to optimize the feature representations of binding residues underlying the imbalanced dataset. We demonstrate that our proposed method significantly outperforms the state-of-the-art methods under benchmarking comparison, and achieves more robust performance. Moreover, we found that we further improve the performance via the integration of traditional features and our learnt features. Our results highlight the flexibility and adaptability of deep learning-based protein language model to capture both conserved and non-conserved sequential characteristics of peptide-binding residues. Interestingly, we demonstrate that peptide-binding residues in local sequential regions have more specific sequential patterns as compared with other protein-ligand binding residues, which potentially provides functional difference. Finally, to facilitate the use of our method, we establish an online predictive platform as the implementation of the proposed PepBCL, which is now available at http://server.wei-group.net/PepBCL/.

AVAILABILITY

https://github.com/Ruheng-W/PepBCL.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Complementarity of the residue-level protein function and structure predictions in human proteins

Sequence-based predictors of the residue-level protein function and structure cover a broad spectrum of characteristics including intrinsic disorder, secondary structure, solvent accessibility and binding to nucleic acids. They were catalogued and evaluated in numerous surveys and assessments. However, methods focusing on a given characteristic are studied separately from predictors of other characteristics, while they are typically used on the same proteins. We fill this void by studying complementarity of a representative collection of methods that target different predictions using a large, taxonomically consistent, and low similarity dataset of human proteins. First, we bridge the gap between the communities that develop structure-trained vs. disorder-trained predictors of binding residues. Motivated by a recent study of the protein-binding residue predictions, we empirically find that combining the structure-trained and disorder-trained predictors of the DNA-binding and RNA-binding residues leads to substantial improvements in predictive quality. Second, we investigate whether diverse predictors generate results that accurately reproduce relations between secondary structure, solvent accessibility, interaction sites, and intrinsic disorder that are present in the experimental data. Our empirical analysis concludes that predictions accurately reflect all combinations of these relations. Altogether, this study provides unique insights that support combining results produced by diverse residue-level predictors of protein function and structure.

Deep learning in prediction of intrinsic disorder in proteins

Intrinsic disorder prediction is an active area that has developed over 100 predictors. We identify and investigate a recent trend towards the development of deep neural network (DNN)-based methods. The first DNN-based method was released in 2013 and since 2019 deep learners account for majority of the new disorder predictors. We find that the 13 currently available DNN-based predictors are diverse in their topologies, sizes of their networks and the inputs that they utilize. We empirically show that the deep learners are statistically more accurate than other types of disorder predictors using the blind test dataset from the recent community assessment of intrinsic disorder predictions (CAID). We also identify several well-rounded DNN-based predictors that are accurate, fast and/or conveniently available. The popularity, favorable predictive performance and architectural flexibility suggest that deep networks are likely to fuel the development of future disordered predictors. Novel hybrid designs of deep networks could be used to adequately accommodate for diversity of types and flavors of intrinsic disorder. We also discuss scarcity of the DNN-based methods for the prediction of disordered binding regions and the need to develop more accurate methods for this prediction.

PIPENN: protein interface prediction from sequence with an ensemble of neural nets

MOTIVATION

The interactions between proteins and other molecules are essential to many biological and cellular processes. Experimental identification of interface residues is a time-consuming, costly and challenging task, while protein sequence data are ubiquitous. Consequently, many computational and machine learning approaches have been developed over the years to predict such interface residues from sequence. However, the effectiveness of different Deep Learning (DL) architectures and learning strategies for protein-protein, protein-nucleotide and protein-small molecule interface prediction has not yet been investigated in great detail. Therefore, we here explore the prediction of protein interface residues using six DL architectures and various learning strategies with sequence-derived input features.

RESULTS

We constructed a large dataset dubbed BioDL, comprising protein-protein interactions from the PDB, and DNA/RNA and small molecule interactions from the BioLip database. We also constructed six DL architectures, and evaluated them on the BioDL benchmarks. This shows that no single architecture performs best on all instances. An ensemble architecture, which combines all six architectures, does consistently achieve peak prediction accuracy. We confirmed these results on the published benchmark set by Zhang and Kurgan (ZK448), and on our own existing curated homo- and heteromeric protein interaction dataset. Our PIPENN sequence-based ensemble predictor outperforms current state-of-the-art sequence-based protein interface predictors on ZK448 on all interaction types, achieving an AUC-ROC of 0.718 for protein-protein, 0.823 for protein-nucleotide and 0.842 for protein-small molecule.

AVAILABILITY AND IMPLEMENTATION

Source code and datasets are available at https://github.com/ibivu/pipenn/.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Computational prediction of protein functional sites-Applications in biotechnology and biomedicine

There are many computational approaches for predicting protein functional sites based on different sequence and structural features. These methods are essential to cope with the sequence deluge that is filling databases with uncharacterized protein sequences. They complement the more expensive and time-consuming experimental approaches by pointing them to possible candidate positions. In many cases they are jointly used to characterize the functional sites in proteins of biotechnological and biomedical interest and eventually modify them for different purposes. There is a clear trend towards approaches based on machine learning and those using structural information, due to the recent developments in these areas. Nevertheless, "classic" methods based on sequence and evolutionary features are still playing an important role as these features are strongly related to functionality. In this review, the main approaches for predicting general functional sites in a protein are discussed, with a focus on sequence-based approaches.

From complete cross-docking to partners identification and binding sites predictions

Proteins ensure their biological functions by interacting with each other. Hence, characterising protein interactions is fundamental for our understanding of the cellular machinery, and for improving medicine and bioengineering. Over the past years, a large body of experimental data has been accumulated on who interacts with whom and in what manner. However, these data are highly heterogeneous and sometimes contradictory, noisy, and biased. Ab initio methods provide a means to a "blind" protein-protein interaction network reconstruction. Here, we report on a molecular cross-docking-based approach for the identification of protein partners. The docking algorithm uses a coarse-grained representation of the protein structures and treats them as rigid bodies. We applied the approach to a few hundred of proteins, in the unbound conformations, and we systematically investigated the influence of several key ingredients, such as the size and quality of the interfaces, and the scoring function. We achieved some significant improvement compared to previous works, and a very high discriminative power on some specific functional classes. We provide a readout of the contributions of shape and physico-chemical complementarity, interface matching, and specificity, in the predictions. In addition, we assessed the ability of the approach to account for protein surface multiple usages, and we compared it with a sequence-based deep learning method. This work may contribute to guiding the exploitation of the large amounts of protein structural models now available toward the discovery of unexpected partners and their complex structure characterisation.

Surveying over 100 predictors of intrinsic disorder in proteins

INTRODUCTION

Intrinsic disorder prediction field develops, assesses, and deploys computational predictors of disorder in protein sequences and constructs and disseminates databases of these predictions. Over 40 years of research resulted in the release of numerous resources.

AREAS COVERED

We identify and briefly summarize the most comprehensive to date collection of over 100 disorder predictors. We focus on their predictive models, availability and predictive performance. We categorize and study them from a historical point of view to highlight informative trends.

EXPERT OPINION

We find a consistent trend of improvements in predictive quality as newer and more advanced predictors are developed. The original focus on machine learning methods has shifted to meta-predictors in early 2010s, followed by a recent transition to deep learning. The use of deep learners will continue in foreseeable future given recent and convincing success of these methods. Moreover, a broad range of resources that facilitate convenient collection of accurate disorder predictions is available to users. They include web servers and standalone programs for disorder prediction, servers that combine prediction of disorder and disorder functions, and large databases of pre-computed predictions. We also point to the need to address the shortage of accurate methods that predict disordered binding regions.

Structure-aware protein-protein interaction site prediction using deep graph convolutional network

MOTIVATION

Protein-protein interactions (PPI) play crucial roles in many biological processes, and identifying PPI sites is an important step for mechanistic understanding of diseases and design of novel drugs. Since experimental approaches for PPI site identification are expensive and time-consuming, many computational methods have been developed as screening tools. However, these methods are mostly based on neighbored features in sequence, and thus limited to capture spatial information.

RESULTS

We propose a deep graph-based framework deep Graph convolutional network for Protein-Protein-Interacting Site prediction (GraphPPIS) for PPI site prediction, where the PPI site prediction problem was converted into a graph node classification task and solved by deep learning using the initial residual and identity mapping techniques. We showed that a deeper architecture (up to eight layers) allows significant performance improvement over other sequence-based and structure-based methods by more than 12.5% and 10.5% on AUPRC and MCC, respectively. Further analyses indicated that the predicted interacting sites by GraphPPIS are more spatially clustered and closer to the native ones even when false-positive predictions are made. The results highlight the importance of capturing spatially neighboring residues for interacting site prediction.

AVAILABILITY AND IMPLEMENTATION

The datasets, the pre-computed features, and the source codes along with the pre-trained models of GraphPPIS are available at https://github.com/biomed-AI/GraphPPIS. The GraphPPIS web server is freely available at https://biomed.nscc-gz.cn/apps/GraphPPIS.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

DeepDISOBind: accurate prediction of RNA-, DNA- and protein-binding intrinsically disordered residues with deep multi-task learning

Proteins with intrinsically disordered regions (IDRs) are common among eukaryotes. Many IDRs interact with nucleic acids and proteins. Annotation of these interactions is supported by computational predictors, but to date, only one tool that predicts interactions with nucleic acids was released, and recent assessments demonstrate that current predictors offer modest levels of accuracy. We have developed DeepDISOBind, an innovative deep multi-task architecture that accurately predicts deoxyribonucleic acid (DNA)-, ribonucleic acid (RNA)- and protein-binding IDRs from protein sequences. DeepDISOBind relies on an information-rich sequence profile that is processed by an innovative multi-task deep neural network, where subsequent layers are gradually specialized to predict interactions with specific partner types. The common input layer links to a layer that differentiates protein- and nucleic acid-binding, which further links to layers that discriminate between DNA and RNA interactions. Empirical tests show that this multi-task design provides statistically significant gains in predictive quality across the three partner types when compared to a single-task design and a representative selection of the existing methods that cover both disorder- and structure-trained tools. Analysis of the predictions on the human proteome reveals that DeepDISOBind predictions can be encoded into protein-level propensities that accurately predict DNA- and RNA-binding proteins and protein hubs. DeepDISOBind is available at https://www.csuligroup.com/DeepDISOBind/.

DNAgenie: accurate prediction of DNA-type-specific binding residues in protein sequences

Efforts to elucidate protein-DNA interactions at the molecular level rely in part on accurate predictions of DNA-binding residues in protein sequences. While there are over a dozen computational predictors of the DNA-binding residues, they are DNA-type agnostic and significantly cross-predict residues that interact with other ligands as DNA binding. We leverage a custom-designed machine learning architecture to introduce DNAgenie, first-of-its-kind predictor of residues that interact with A-DNA, B-DNA and single-stranded DNA. DNAgenie uses a comprehensive physiochemical profile extracted from an input protein sequence and implements a two-step refinement process to provide accurate predictions and to minimize the cross-predictions. Comparative tests on an independent test dataset demonstrate that DNAgenie outperforms the current methods that we adapt to predict residue-level interactions with the three DNA types. Further analysis finds that the use of the second (refinement) step leads to a substantial reduction in the cross predictions. Empirical tests show that DNAgenie's outputs that are converted to coarse-grained protein-level predictions compare favorably against recent tools that predict which DNA-binding proteins interact with double-stranded versus single-stranded DNAs. Moreover, predictions from the sequences of the whole human proteome reveal that the results produced by DNAgenie substantially overlap with the known DNA-binding proteins while also including promising leads for several hundred previously unknown putative DNA binders. These results suggest that DNAgenie is a valuable tool for the sequence-based characterization of protein functions. The DNAgenie's webserver is available at http://biomine.cs.vcu.edu/servers/DNAgenie/.

dSPRINT: predicting DNA, RNA, ion, peptide and small molecule interaction sites within protein domains

Domains are instrumental in facilitating protein interactions with DNA, RNA, small molecules, ions and peptides. Identifying ligand-binding domains within sequences is a critical step in protein function annotation, and the ligand-binding properties of proteins are frequently analyzed based upon whether they contain one of these domains. To date, however, knowledge of whether and how protein domains interact with ligands has been limited to domains that have been observed in co-crystal structures; this leaves approximately two-thirds of human protein domain families uncharacterized with respect to whether and how they bind DNA, RNA, small molecules, ions and peptides. To fill this gap, we introduce dSPRINT, a novel ensemble machine learning method for predicting whether a domain binds DNA, RNA, small molecules, ions or peptides, along with the positions within it that participate in these types of interactions. In stringent cross-validation testing, we demonstrate that dSPRINT has an excellent performance in uncovering ligand-binding positions and domains. We also apply dSPRINT to newly characterize the molecular functions of domains of unknown function. dSPRINT's predictions can be transferred from domains to sequences, enabling predictions about the ligand-binding properties of 95% of human genes. The dSPRINT framework and its predictions for 6503 human protein domains are freely available at http://protdomain.princeton.edu/dsprint.

GraphBind: protein structural context embedded rules learned by hierarchical graph neural networks for recognizing nucleic-acid-binding residues

Knowledge of the interactions between proteins and nucleic acids is the basis of understanding various biological activities and designing new drugs. How to accurately identify the nucleic-acid-binding residues remains a challenging task. In this paper, we propose an accurate predictor, GraphBind, for identifying nucleic-acid-binding residues on proteins based on an end-to-end graph neural network. Considering that binding sites often behave in highly conservative patterns on local tertiary structures, we first construct graphs based on the structural contexts of target residues and their spatial neighborhood. Then, hierarchical graph neural networks (HGNNs) are used to embed the latent local patterns of structural and bio-physicochemical characteristics for binding residue recognition. We comprehensively evaluate GraphBind on DNA/RNA benchmark datasets. The results demonstrate the superior performance of GraphBind than state-of-the-art methods. Moreover, GraphBind is extended to other ligand-binding residue prediction to verify its generalization capability. Web server of GraphBind is freely available at http://www.csbio.sjtu.edu.cn/bioinf/GraphBind/.

DELPHI: accurate deep ensemble model for protein interaction sites prediction

MOTIVATION

Proteins usually perform their functions by interacting with other proteins, which is why accurately predicting protein-protein interaction (PPI) binding sites is a fundamental problem. Experimental methods are slow and expensive. Therefore, great efforts are being made towards increasing the performance of computational methods.

RESULTS

We propose DEep Learning Prediction of Highly probable protein Interaction sites (DELPHI), a new sequence-based deep learning suite for PPI-binding sites prediction. DELPHI has an ensemble structure which combines a CNN and a RNN component with fine tuning technique. Three novel features, HSP, position information and ProtVec are used in addition to nine existing ones. We comprehensively compare DELPHI to nine state-of-the-art programmes on five datasets, and DELPHI outperforms the competing methods in all metrics even though its training dataset shares the least similarities with the testing datasets. In the most important metrics, AUPRC and MCC, it surpasses the second best programmes by as much as 18.5% and 27.7%, respectively. We also demonstrated that the improvement is essentially due to using the ensemble model and, especially, the three new features. Using DELPHI it is shown that there is a strong correlation with protein-binding residues (PBRs) and sites with strong evolutionary conservation. In addition, DELPHI's predicted PBR sites closely match known data from Pfam. DELPHI is available as open-sourced standalone software and web server.

AVAILABILITY AND IMPLEMENTATION

The DELPHI web server can be found at delphi.csd.uwo.ca/, with all datasets and results in this study. The trained models, the DELPHI standalone source code, and the feature computation pipeline are freely available at github.com/lucian-ilie/DELPHI.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Assessment of protein-protein interfaces in cryo-EM derived assemblies

Structures of macromolecular assemblies derived from cryo-EM maps often contain errors that become more abundant with decreasing resolution. Despite efforts in the cryo-EM community to develop metrics for map and atomistic model validation, thus far, no specific scoring metrics have been applied systematically to assess the interface between the assembly subunits. Here, we comprehensively assessed protein-protein interfaces in macromolecular assemblies derived by cryo-EM. To this end, we developed Protein Interface-score (PI-score), a density-independent machine learning-based metric, trained using the features of protein-protein interfaces in crystal structures. We evaluated 5873 interfaces in 1053 PDB-deposited cryo-EM models (including SARS-CoV-2 complexes), as well as the models submitted to CASP13 cryo-EM targets and the EM model challenge. We further inspected the interfaces associated with low-scores and found that some of those, especially in intermediate-to-low resolution (worse than 4 Å) structures, were not captured by density-based assessment scores. A combined score incorporating PI-score and fit-to-density score showed discriminatory power, allowing our method to provide a powerful complementary assessment tool for the ever-increasing number of complexes solved by cryo-EM.