Abstract
ProteinProtein Interactions (PPIs) are fundamental means of functions and signalings in biological systems. The massive growth in demand and cost associated with experimental PPI studies calls for computational tools for automated prediction and understanding of PPIs. Despite recent progress, in silico methods remain inadequate in modeling the natural PPI hierarchy. Here we present a doubleviewed hierarchical graph learning model, HIGHPPI, to predict PPIs and extrapolate the molecular details involved. In this model, we create a hierarchical graph, in which a node in the PPI network (top outsideofprotein view) is a protein graph (bottom insideofprotein view). In the bottom view, a group of chemically relevant descriptors, instead of the protein sequences, are used to better capture the structurefunction relationship of the protein. HIGHPPI examines both outsideofprotein and insideofprotein of the human interactome to establish a robust machine understanding of PPIs. This model demonstrates high accuracy and robustness in predicting PPIs. Moreover, HIGHPPI can interpret the modes of action of PPIs by identifying important binding and catalytic sites precisely. Overall, “HIGHPPI [https://github.com/zqgao22/HIGHPPI]” is a domainknowledgedriven and interpretable framework for PPI prediction studies.
Similar content being viewed by others
Introduction
Biological functions are accomplished by interactions and chemical reactions among biomolecules. Among them, protein–protein interactions (PPIs) are arguably one of the most important molecular events in the human body and are an important source of therapeutic interventions against diseases. A comprehensive dictionary of PPIs can help connect the dots in complicated biological pathways and expedite the development of therapeutic^{1,2}. In biology, hierarchy information has been widely exploited to gain indepth information about phenotypes of interest, for example, in disease biology^{3,4,5}, proteomics^{6,7,8}, and neurobiology^{9,10,11}. Naturally, PPIs encapsulate a twoview hierarchy: on the top view, proteins interact with each other; on the bottom view, key amino acids or residues assemble to form important local domains. Following this logic, biologists often take hierarchical approaches to understand PPIs^{12,13}. Experimentally, scientists often employ highthroughput mapping^{14,15,16} to prebuild the PPI network at scale, and use bioinformatics clustering methods to identify functional modules of the network (top view). On the individual protein level, isolation methods, such as coimmunoprecipitation^{17}, pulldown^{18}, and crosslinking^{19} are used to establish the structures of individual proteins, so that surficial ‘hotspots’ can be located and analyzed. In short, hierarchy knowledge of structure information is important to understand the molecular details of PPIs.
More recently, the massive growth in the demand and the cost of experimentally validating PPIs make it impossible to characterize most unknown PPIs in wet laboratories. To map out the human interactome efficiently and inexpensively, computational methods are increasingly being used to predict PPIs automatically. Over the past decade, as one of the most revolutionary tools in computation, Deep Learning (DL) methods, have been applied to study PPIs. Development in this field has been mostly focused on two aspects, learning appropriate protein representations^{20,21} and inferring potential PPIs by link predictions^{22,23}. The former focuses on extracting structural information using protein sequences. In particular, Convolutional Neural Networks (CNNs)^{24,25} and Recurrent Neural Networks (RNNs)^{26,27,28} have demonstrated high generalization and fast inference speed to capture key sequence fragments for PPIs^{29}. 3D CNNs^{21,30,31} have shown to be better at extracting 3D structural features of proteins and thus capturing the spatialbiological arrangements of residues^{32} that are important to PPI predictions. However, 3D CNN suffers from high computational burdens and limited resolution that is prone to quantization errors^{29}. The latter aspect of DL in PPI predictions focuses on the PPI network structures, which involves developing link prediction methods to identify missing interactions within the known network topology. Link prediction methods based on common neighbor (CN)^{33} assign high probabilities of PPI to protein pairs that are known to share common PPI partners. CN can be generalized to consider neighbors from a greater path length (L3)^{22}, which captures the structural and evolutionary forces that govern biological networks such as the interactome. Additionally, distancebased methods measure the possible distances between protein pairs, such as Euclidean commute time (ECT)^{34} and random walk with restart (RWR)^{35}. Most methods of traditional link prediction focus on known interactions but tend to overlook important network properties such as node degrees and community partitions.
More importantly, these methods perceive only one of the two views of outsideofprotein and insideofprotein. Few can model the natural PPI hierarchy by connecting both views. To address this issue, we present a hierarchical graph that applies two Graph Neural Networks (GNNs)^{36,37} to represent protein and network structures, respectively. In this way, the limitations of 3D CNN and link prediction methods mentioned above can be circumvented. First, GNNs can learn the protein 3D structures on more efficient graph representations, even when facing highresolution requirements for structure processing. Second, due to the propagation mechanism, GNNs are capable of recovering network properties such as node degrees and community partitions. In short, this hierarchical graph approach aims at modeling the natural PPI hierarchy with more effective and efficient structure perceptions.
Here we describe a generic DL platform tailored for predicting PPIs, Hierarchical Graph Neural Networks for Protein–Protein Interactions (HIGHPPI). HIGHPPI models the structural protein representations with the bottom insideofprotein view GNNs (BGNN) and the PPI network with the top outsideofprotein view GNNs (TGNN). In the bottom view, HIGHPPI constructs protein graphs by treating amino acid residues as nodes and physical adjacencies as edges. Thus, BGNN integrates the information of protein 3D structures and residuelevel properties in a synergistic fashion. In the top view, HIGHPPI constructs the PPI graph by taking protein graphs (the bottom view) as nodes and interactions as edges and learns protein–protein relationships with TGNN. In an endtoend training paradigm, HIGHPPI gains mutual benefits from both views. On the one hand, the bottom view feeds protein representations to the top view to learn accurate protein relationships. On the other hand, protein relationships learned by the top view provide insights to further optimize the bottom view to establish better protein representations. HIGHPPI outputs the probabilities of interactions for given protein pairs and predicts key “contact” sites for such interactions by calculating residue importance. We show the effectiveness of HIGHPPI on the human interactome from the STRING database^{38} and compare it with leading DL methods. We demonstrate the superiority of HIGHPPI with higher prediction accuracy and better interpretability. We also show examples that HIGHPPI can identify binding and catalytic sites with high precision automatically.
Results
HIGHPPI introduces a hierarchical graph for learning structures of proteins and the PPI network
Although deep learning (DL) models for Protein–Protein Interaction (PPI) prediction have been studied extensively, it has not yet been developed for simulating the natural PPI hierarchy. Here, we suggest HIGHPPI, a hierarchical graph neural network, for accurate and interpretable PPI prediction. HIGHPPI works like biologists in a hierarchical manner as it contains the bottom insideofprotein view and top outsideofprotein view (schematic view in Fig. 1c and detailed architecture in Supplementary Fig. 1a). On one hand, HIGHPPI applies the bottom view when dealing with a protein, where a protein is represented by a protein graph with residue as nodes and their physical adjacencies as edges. On the other hand, from the top view, protein graphs and their interactions are considered nodes and edges of the PPI graph, respectively. Correspondingly, two GNNs are respectively employed to learn from protein graphs in the bottom view (BGNN) and learn from a PPI graph in the top view (TGNN). Consequently, a set of graphs are interconnected by edges in a hierarchical graph, to present a potent data representation.
In the proposed endtoend model, the initial stage is to create protein graphs for learning appropriate protein representation. An adjacency matrix of a protein graph is derived from a contact map connecting physically close residues (See Section 4.1 in “Methods” for details). Node attributes are defined with residuelevel features for expressing the physicochemical properties of proteins (See Section 4.1 in “Methods” for details). To produce a protein graph representation, Graph Convolutional Network (GCN)^{36} is used in BGNN to optimize the protein graphs. As shown in Fig. 1c, BGNN contains two GCN blocks, and we construct three components for each GCN block to obtain a fixedlength embedding vector for a protein graph. Both the adjacency matrix and the residuelevel features matrix are inputs for a GCN layer. To respectively improve model expressiveness and accelerate training convergence, the nonlinear activation function of ReLU and Batch Normalization (BN) are used. Readout operation including a selfattention graph (SAG) pooling^{39} and the average aggregation is used to ensure a fixedlength embedding vector output. Regardless of the number and permutation of residues, a 1D embedding vector is obtained after two GCN blocks. By the end of those operations, the final protein representations are assembled, which are employed as initial features of the PPI graph. In TGNN, features are propagated along interactions in the PPI network for learning network community and degree properties. In the top view, we specifically design a GIN block that contains a Graph Isomorphism Network (GIN)^{37} layer, ReLU activation function and a BN layer. Node features of the PPI graph are updated with recursive neighborhood aggregations of three GIN blocks. Two arbitrary protein embeddings are combined by concatenation operations, and a MultiLayer Perceptron (MLP) is then applied as a classifier for prediction. Moreover, we also consider graph attention (GAT) and arbitrarily deploy two of the three GNN layers (i.e., GCN, GIN and GAT) on BGNN and TGNN. The performance of HIGHPPI with various GNN layers is shown in Supplementary Fig. 2.
We train and evaluate HIGHPPI on multitype human PPIs from the STRING database^{38}, which contains a critical assessment and integration of PPIs. SHS27k^{26}, a homo sapiens subset from STRING^{38} that comprises 1,690 proteins and 7,624 PPIs, is used to train and evaluate the HIGHPPI unless otherwise noted. However, a small fraction of proteins (∼ 8%) sometimes need to be removed because of the lack of their native structures in the PDB database. While evaluating the prediction performance for multitype PPIs, we consider the prediction for each PPI type as a onevsall binary classification problem, for which two metrics, F1 score and area under the precisionrecall curve (AUPR) are used for predicting the presence or absence of the corresponding PPI class. The overall performance of microF1 and AUPR scores for multitype PPI prediction is averaged across all PPI types.
HIGHPPI shows the best performance, robustness and generalization
To validate the predictive power of our model, we compare HIGHPPI with leading methods from four perspectives, including (1) the overall performance under a random data split, (2) the robustness of HIGHPPI against random interaction perturbation, (3) model generalization for predicting PPI pairs containing unknown proteins, (4) evaluations in terms of AUPR on five separate PPI types. For each method, all the proposed modules and strategies are involved to get the best performance.
First, we compare the overall performance of HIGHPPI with leading baselines in Fig. 2a. To ensure native PDB structures for all proteins, we filter from SHS27k and construct the dataset containing ∼1600 proteins (see Supplementary Data File 1) and ∼6600 PPIs. We randomly select 20% PPIs for testing and compare PPI to one stateoftheart DL method (i.e., GNNPPI^{24}), one sequencebased method (i.e., PIPR^{26}), one 2D CNNbased method (i.e., DrugVQA^{40}) and one machine learning (ML) method based on random forest (i.e., RFPPI^{41}). GNNPPI applies a GNN module to learn the PPI network topology and 1D CNN to learn protein representations by taking pretrained residue embeddings as inputs. PIPR, an endtoend framework based on recurrent neural networks (RNN), represents proteins with only pretrained residue embeddings. DrugVQA applies a visual questionanswering mode to learn from protein contact maps with a 2D CNN model and extract semantic features with a sequential model. Supplementary Data File 2 contains predictions of HIGHPPI for all test PPIs from SHS27k. We provide the precisionrecall curves in Fig. 2a. In terms of best microF1 scores (bestF1), HIGHPPI obtains the best performance. Pretrained residue embedding method GNNPPI takes the second place by effectively generalizing to unknown proteins. Without using any pretraining techniques, HIGHPPI surpasses GNNPPI by an average of ∼4%, showing the superiority of the hierarchical modeling approach. DrugVQA gets relatively poor performance (bestF1 ≈ 0.7), which could be attributed to the neglect of residue property information and structures of the PPI network.
Second, to evaluate the robustness of HIGHPPI, we analyze the model tolerance against interaction data perturbation including random addition or removal of known interactions. This simulates scenarios where PPI datasets always omit undiscovered interactions and may introduce mislabeled ones. Based on the perturbated PPI network, we split the training and test sets at an 8:2 ratio. We observe in Fig. 2b that our method exhibits stable performance in terms of bestF1 with a random perturbation of 40%. When compared to the secondbest baseline (i.e., GNNPPI), HIGHPPI offers a significant performance gain of up to 19%, which demonstrates the strongest model robustness among all methods. It is crucial to notice that although RFPPI and DrugVQA perform consistently in the overall evaluation (see Fig. 2a), DrugVQA performs significantly more robustly than RFPPI, demonstrating the undisputed superiority of DL methods over ML ones. Furthermore, we perform false discovery on our method, which investigates the effect of the training data unreliability (i.e., false negative (FN) and false positive (FP)) on our model and a solid baseline (GNNPPI). Specifically, we consider the original dataset to be reliable and artificially add perturbations to represent data unreliability. Supplementary Table 1 shows the created 9 datasets with different FP rates (\({{FPR}}_{{train}}\)) and FN rates (\({{FNR}}_{{train}}\)). We respectively train the model on the reliable training set and created 9 unreliable ones and present the FP rates (\({{FPR}}_{{pre}}\)), FN rates (\({{FNR}}_{{pre}}\)) and false discovery rates (\({{FDR}}_{{pre}}\)) metrics on the test sets (see Supplementary Table 2 and 3). Without unreliability, our model achieves best performance with insignificant superiority (*\(P=3.8\times {10}^{2}\)) in the \({{FPR}}_{{pre}}\) metric, and considerable superiority in the \({{FNR}}_{{pre}}\) (***\(P=1.2\times {10}^{4}\)) and \({{FDR}}_{{pre}}\) (***\(P=1.5\times {10}^{4}\)) metrics. When introducing data unreliability, we are surprised to find that our model substantially improves the superiority significance in the \({{FPR}}_{{pre}}\) metric (****\(P=4.0\times {10}^{5}\)) while retaining the original significance in \({{FNR}}_{{pre}}\) and \({{FDR}}_{{pre}}\). In addition to showing the excellent robustness of our model, we also provide more indepth insights in Section 3.2.
Generalization ability is investigated by testing HIGHPPI in various outofdistribution (OOD) scenarios where unknown proteins arrive in the test sets with different probabilities (see Fig. 2c). For example, BFS0.3 denotes that the test set involves 30% known proteins via BreathFirst Search approach^{24}. For PIPR, DrugVQA and RFPPI, we visualize their best performances among all OOD cases using dotted lines, to demonstrate the absolute dominance of HIGHPPI and GNNPPI. Furthermore, we observe that HIGHPPI consistently outperforms GNNPPI, the secondbest method, with large margins in all five scenarios. BFS typically produces worse performance than DFS, because BFS creates a more challenging and realistic mode where unknown proteins exist in cluster forms. ML method (RFPPI) exhibits poor generalization. Furthermore, we follow Park and Marcotte^{42} to explore the differences in model performance on 3 kinds of PPI pairs with different degrees of OOD. Specifically, \({C}_{1}\) stands for the percentage of PPIs of which both proteins were present in a training set (Class 1), \({C}_{2}\) stands for the percentage of PPIs of which one of (but not both) proteins was present in the training set (Class 2), \({C}_{3}\) stands for the percentage of PPIs of which neither protein was present in the training set (Class 3). The detailed experimental protocol has been presented in the Supplementary Method 3. We come to the same conclusion as Park and Marcotte did^{42}. There is a noticeable difference in model test performance across the 3 distinct classes of test pairs. Particularly, on Class 1 test pairs, both models (HIGHPPI and GNNPPI) perform the best, on Class 2 test pairs they are the second best, and on Class 3 test pairs they are the poorest. Furthermore, we find that for each model, the class proportion (i.e., \({C}_{1}/{C}_{2}/{C}_{3}\)) had an impact on the overall performance of the model despite having little effect on performance on the respective classes. Thus, it seems that the proportion of the three test pair classes (Supplementary Table 6) as well as the percentage of unknown proteins (Fig. 2c) in the test sets may both have a significant role in determining the degree of OOD in the dataset.
Finally, for each of the five PPI types, we offer a separate performance analysis in terms of AUPR. In all five types, HIGHPPI consistently beats other baselines with high significance as shown in Fig. 2d. As anticipated, PPI types with high proportions (such as binding, reaction, and catalysis) can be predicted more easily since the model could learn enough relevant information. In addition, we find that when predicting binding PPIs, HIGHPPI outperforms GNNPPI most significantly (****\(P=2.0\times {10}^{5}\)). This is reasonable as HIHGPPI is designed to recognize spatialbiological patterns of proteins, which is highly related to binding type PPIs. Similar trends are also found in the performance of HIGHPPI and GNNPPI in various PPI types under OOD cases (Supplementary Fig. 5).
Bottom insideofprotein view improves the performance
We investigate the role of the bottom insideofprotein view from four perspectives, including (1) the effectiveness of graph representations and backbones with native protein structures, (2) the model tolerance with lowquality protein structures, (3) the capability to predict motifs (i.e., functional sites) in a protein, (4) the overall and typespecific feature importance.
First, we explore the effectiveness of backbones including RF, RNN, CNN and GNN in Fig. 3a. For fairness, we feed the same features of residue sequence to RF, RNN and CNN, whose results are displayed by bar charts with ‘Seq’. We directly use RFPPI as the RF backbone. For RNN and CNN backbones, we respectively employ the RNN module of PIPR and the CNN module of GNNPPI to extract sequence embeddings for representing proteins and apply the same fully connected layer as classifiers. We test the predictive power of each model with 3D information. For RF and RNN, we employ the concatenations of sequence data and Cartesian 3D coordinates of each \({C}_{\alpha }\). For CNN, we apply the 3D CNN module suggested in DeepRank^{21}, a deep learning framework for identifying interfaces of PPIs. For GNN, we learn from protein graphs in which the adjacency matrix is determined by \({C}_{\alpha }{C}_{\alpha }\) contact map. With the aid of 3D information, we discover all the model performance can be improved, indicating that 3D information is an important complement to sequencealone information. Importantly, GNN performs the best when compared to RF ( + 3D), RNN ( + 3D) and CNN ( + 3D), which shows that GNN is the best approach for capturing spatialbiological arrangements of residues within a protein. Moreover, GNN performs significantly better than 3D CNN in memory and time efficiency (Supplementary Fig. 3).
Second, we examine the model tolerance when testing with lowquality structure data (see Fig. 3b). This meets the realistic scenarios, where native structure information is not always available for predicting PPIs. We prefer the model whose performance is not seriously limited by the structure quality, which is robust to inputs directly from computational models (e.g., AlphaFold^{43}). We evaluate the quality of the input protein structure by calculating the rootmeansquare deviation (RMSD) of the native one and the input. Native protein structures (RMSD = 0) are retrieved from the PDB database at the highest resolutions. We compute the bestF1 scores (box plots) of our method on a set of AlphaFold structures with various RMSDs (0.80, 1.59, 2.39, 3.19, 5.36, 7.98), and show the average result of secondbest method (GNNPPI) in a blue dotted line. As can be seen, our model performance is always better than GNNPPI, even with RMSD up to 8. The comparison with 3D CNN model^{21} further proves the denoising ability of the hierarchical graph for protein structure errors (Supplementary Fig. 4a). In short, our model performance is not significantly affected by structure errors where powerful pretrained features are not available.
Further, to interpret decisions made by RNN, CNN and GNN, an experiment is conducted to explore the ability to capture protein functional sites. We apply the 3DgradCAM approach^{44} on the trained 3D CNN model named DeepRank^{21}, and apply the RNNVis approach^{45} on the trained PIPR^{26} model with 3D information. All three methods have identified more than one motif, in which we only show the most crucial site. Figure 3c displays the binding site for an isomerase protein’s chain A (PDB id: 1BJP). The binding site is made up of four residues with the sequence numbers 6, 42, 43, and 44. As can be seen, whereas neither CNN nor RNN can identify the His6 residue, our method can precisely identify the binding site by using graph motif search. It seems to be a challenge for the sequence model (i.e., RNN, CNN) to connect His6 to the other residues, probably because of their weak connections in a sequence mode. Moreover, 3D CNN performs even worse than RNN as it incorrectly classifies the nonessential Ile41 residue.
For node features in protein graphs, we select seven important features from twelve residuelevel feature options (see Supplementary Table 4) that are easily available. The feature selection process (see Supplementary Method 1 for details) produces the optimal set consisting of seven features to ensure that our model peaks at both AUPR and bestF1 scores. Here, we list the selected seven residuelevel physicochemical properties in Fig. 3d and discuss their importance for different types of PPIs to both better interpret our model and discover enlightening biomarkers for PPI interface. The average zscore, which results from deleting each feature dimension and analyzing changes in AUPR before and after, is calculated to determine the importance of a feature. We choose a representative type (i.e., binding) to explain because it is the most prevalent in the STRING database. As a consequence, HIGHPPI regards topological polar surface area (TPSA) and octanolwater partition coefficient (KOW) as dominant features. This finding supports the conventional wisdom that TPSA and KOW play a key role in drug transport process^{46}, protein interface recognition^{47,48}, and PPI prediction^{49}.
Top outsideofprotein view improves the performance
We investigate the role of top outsideofprotein view TGNN from three perspectives, including (1) the importance of degree and community recovery for predicting network structures, (2) comparison results of TGNN and other leading link prediction methods, (3) a reallife example to show the shortcomings of the leading link prediction methods.
Recently, various works have demonstrated the usefulness of structure properties (e.g., degree, community) of networks for predicting missing links. HIGHPPI is inspired to efficiently recover the degree and community partitions of the PPI network by utilizing the network topology. We show an empirical study in Fig. 4a to illustrate the impact of degree and community recovery for link prediction. We randomly select the test results from the model trained in different epochs and calculate the negative Mean Absolute Error (MAE) of the predicted degrees and real degrees to represent degree recovery. Similarly, for community recovery, we quantify the community recovery using the normalized mutual information (NMI). As can be seen, we observe a significant correlation (\(R=0.66\)) between degree recovery and model performance (i.e., bestF1) as well as a high correlation (\(R=0.68\)) between community recovery and model performance, which means better recovery of the degree and community of PPI network implies better PPI prediction performance.
Second, we evaluate the performance of TGNN and leading link prediction methods using PPI network structure as input. Our method (TGNN) takes interactions as edges and node degrees as node features. We compare HIGHPPI with six heuristic methods and one DLbased method. Heuristic methods, the simple yet effective ones utilizing the heuristic node similarities as the link likelihoods, include common neighbors (CN)^{33}, Katz index (Katz)^{50}, AdamicAdar (AA)^{51}, preferential attachment (PA)^{52}, SimRank (SR)^{53} and paths of length three (L3)^{22}. MLP_IP, a DL approach, learns node representations using a multilayer perceptron (MLP) and identifies the node similarity via inner product (IP) operation. We calculate the MAE and NMI values of recovered networks and highlight those with a high capacity for recovery (NMI ≥ 0.7 and MAE ≤ 0.35) in orange. Results show that link prediction methods that are more adept at recovering network properties typically perform better. This gain validates our findings in Fig. 4a and highlights the need for TGNN in the top view. In addition, a comparison of MIL_IP and L3 elucidates that pairwise learning is insufficient to well capture the network information. Although L3 can capture the evolutionary principles of PPIs to some extent, our method beats L3 by better recovering the structure of the PPI network.
We provide an example on an SHS27k subnetwork. As can be seen, there exist two distinct communities connected by two intercommunity edges. We use the original subnetwork as inputs and find that nonTGNN link prediction methods (i.e., CN, Katz, SR, AA, PA) tend to give high scores for intercommunity interactions. As an interesting observation, when we apply the Louvain community detection algorithm^{54} to the recovered structure, it cannot produce an accurate community partition as the abundant intercommunity interactions disrupt the original community structure. To examine degree recovery ability, we randomly select 50% of interactions as inputs and show each method’s degree recovery result for node KIF22 in Fig. 4c. We find nonTGNN approaches cannot well recover the links connecting the node KIF22 while TGNN approach can. In short, these experiments demonstrate that the structure properties of the PPI network are not always reflected in traditional link prediction methods, and moreover, capturing and learning the network structures in our top view improves the prediction performance.
HIGHPPI accurately identifies key residues constituting functional sites
Typically, functional sites are spatially clustered sets of residues. They control protein functions and are thus important for PPI prediction. As our proposed model has the capacity to capture spatialbiological arrangements of residues in the bottom view, this characteristic can be used to explain the model’s decision. It is meaningful to notice that HIGHPPI can automatically learn the residue importance without any residuelevel annotations. In this section, we provide (1) a case study of predicting residue importance for the binding surface, (2) two cases of estimating residue importance for catalytic sites, and (3) an explainable ability comparison of precision in predicting binding sites.
First, a binding example between the query protein (PDB id: 2B6HA) and its partner (PDB id: 2REYA) is investigated. The ground truth binding surface is retrieved from the PDBePISA database^{55}, which is colored in red in Fig. 5a. Subsequently, we apply the GNN explanation approach (see Section 4.5 in “Methods” for details) on the HIGHPPI model. As can be seen from Fig. 5a, HIGHPPI can accurately and automatically identify the residues belonging to the binding surface. Another observation is shown in Fig. 5c which indicates our learned residue importance is quite close to the real profiles. We show another six cases of HIGHPPI for identifying binding surfaces correctly in Supplementary Fig. 7.
Second, in order to evaluate the prediction of catalytic sites for PPIs, we utilize the same GNN explanation approach in our model. The ground truth catalytic site is retrieved from the Catalytic Site Atlas^{56} (CSA), a database for catalytic residue annotation for enzymes. We calculate the residue importance of catalytic sites for query proteins (PDB id: 1S9IA, 1I0OA). As seen in Fig. 5b, our proposed HIGHPPI can correctly predict both residues for 1S9IA and two out of three for 1I0OA. We show another nine cases of HIGHPPI for identifying catalytic sites in Supplementary Fig. 6, where a total of 25 out of 34 catalytic sites are correctly identified.
Additionally, we compare the model interpretability of the CNN, 3D CNN and HIGIPPI models. We employ the CNN module in GNNPPI^{24} and 3D CNN module in DeepRank^{21}, respectively. We apply gradCAM^{57} and 3DgradCAM^{44} approaches to determine residue importance for CNN and 3D CNN models, correspondingly. We use the binding type PPIs from the STRING dataset as the training set, and randomly select 20 binding type PPIs as the test set. We use the ground truth from PDBePISA for each query protein and treat its residues with importance >0 as surface compositions. To gauge the precision of the surface prediction, intersection over union (IoU) is used, and the box plots of the IoU score distributions are shown in Fig. 5d. The results elucidate that HIGHPPI significantly outperforms other models in terms of interpretability with a minimum variance. In addition, 3D CNN outperforms CNN with a smaller variance, showing that 3D information supports the learning of reliable and generalized protein representations.
Protein functional site prediction sheds light on the model decisions and how to carry out additional experimental validations for PPI investigation. Excellent model interpretability also shows that our approach can accurately describe biological evidence for proteins.
Discussion
Hierarchical graph learning
In this paper, we study the PPI problem from a hierarchical graph perspective and develop a hierarchical graph learning model named HIGHPPI to predict PPIs. Empirically, HIGHPPI for PPI prediction outperforms leading methods by a significant margin. The hierarchical graph exhibits high generalization for recognizing unknown proteins and robustness against protein structure errors and PPI network perturbations.
Even without explicit supervision from binding site information, HIGHPPI demonstrates its ability to capture residue importance for PPI with the aid of a hierarchical graph, which is a good indicator of excellent interpretability. Suppose HIGHPPI predicts the presence of a catalytic interaction for a protein pair but identifies important sites unrelated to catalysis, we will hardly trust the model’s decision. Moreover, interpretability provides trusted guides for subsequent wet experimental validations. For example, if HIGHPPI thinks a catalytic site is important, experiments may be designed by targeting the specific site for validation.
In conclusion, interpretable, endtoend learning with a hierarchical graph revealing the PPI nature can pave the way to map out human interactome and deepen our understanding of PPI mechanisms.
Limitations and future work
We describe our intuitions in the hierarchical graph learning for PPIs. The world is hierarchical. Humans tend to solve problems or learn knowledge by conceptualizing the world from a hierarchical view^{58}. Due to huge semantic gaps between hierarchical views, humans always use a multiview learning strategy to deepen the understanding of one view from the other one. Given rich hierarchical information, recent machine intelligence methods can effectively learn knowledge in each separate view but are not experts in gaining mutual benefits from both views. This is the challenge that our hierarchical world presents to machine intelligence. Here we connect both views by employing the forward and backward propagation of DL models. The forward propagation benefits the learning for the PPI network in the top view. In turn, the backward propagation optimizes the PPIappropriate protein representations in the bottom view.
We describe two main limitations of HIGHPPI and outline potential solutions in future work. (1) We did not explore in depth how to use proteinlevel annotations. Annotations for protein functions are becoming more available due to the recent growth of protein function databases (e.g., the UniProt Knowledgebase^{59}) and computational methods^{29} for protein function prediction. Some annotations may speed up learning PPIs. For example, two proteins with low scores of the “protein binding” function term hardly interact with each other. We suggest that future work may consider leveraging function annotations to enhance the expressiveness of protein representations. Inspired by the contrastive learning principle, a potentially feasible solution is to enhance the consistency in protein representations and functions. (2) Protein domain information may be beneficial for hierarchical models. We clarify the core ideas here and provide a detailed description in Supplementary Method 2. Domains are distinct functional or structural units in proteins and are responsible for PPIs and specific protein functions. Both in terms of structures and functions, the protein domain can represent a crucial middle scale for the PPI hierarchy. However, to our knowledge, true (native) domain annotations are not easily available and predicted ones are usually retrieved from computational tools, which inevitably leads to data unreliability. If we employ the domain scale as a separate view, data unreliability may spread to other views and impair the entire hierarchical model. On this basis, we prefer to recommend domain annotations as supervised information at the residue level. Precisely, a welldesigned regularization is required to guarantee that all functional sites, discovered by HIGHPPI, belong in the prepared domain database. The domain regularization and the PPI prediction loss form a flexible tradeoff of learning objectives, which can appropriately tolerate the domain annotation unreliability. (3) Memory requirement grows with the view number of a hierarchical graph. HIGHPPI employs two views to form the hierarchical graph and treat amino acid residues as microscopic components of proteins. However, we did not further consider one more microscopic view where atoms, the components of residues, provide information for representing residues. It might be beneficial to introduce an atomlevel view and develop a memoryefficient way for storing and processing explicit 3D atomlevel information. (4) In future work, model robustness can be further improved. Although our model outperforms in the robustness evaluation (see Supplementary Table 3), we observe that \({{FDR}}_{{pre}}\) is most impacted by unreliable data, which is mostly because the number of FP significantly increases (up to 6 times) from Data 1 to Data 9. A possible explanation for the significant rise in FP is that the model’s “low demand” for a positive sample permits certain controversial samples to be projected as true. To address this issue, we recommend the future work consider a straightforward method—the voting strategy which uses the voting outcomes of various independent classifiers to identify true PPIs. Independence makes it unlikely for voting classifiers to commit the same errors. A test pair can only be predicted as true if it is approved by most voting classifiers, which makes the model more demanding for the PPI presence.
Methods
Construction of a hierarchical graph
We denote a set of amino acid residues in a protein as \({Prot}=\{{r}_{1},{r}_{2},\ldots,{r}_{n}\}\). Each residue is described with \(\theta\) kinds of physicochemical properties. For the bottom insideofprotein view, a protein graph \({g}_{b}=({V}_{b},{A}_{b},{X}_{b})\) is constructed to model the relationship between residues in \({Prot}\), where \({V}_{b}\subseteq {Prot}\) is the set of nodes, \({A}_{b}\) is an \(n\times n\) adjacency matrix representing the connectivity in \({g}_{b}\), and \({X}_{b}\in {{\mathbb{R}}}^{n\times \theta }\) is a feature matrix containing the properties of all residues.
For the top outsideofprotein view, a set of protein graphs can be interconnected within a PPI graph \({g}_{t}\), which is denoted as \({g}_{b}\in {V}_{t}\). The connectivity (i.e., interactions) between protein graphs can be denoted as an \(m\times m\) adjacency matrix \({A}_{t}\). In addition, \({X}_{t}\in {{\mathbb{R}}}^{m\times \varnothing }\) represents a feature matrix containing the representations of all proteins. We model the protein graphs and their connections as a hierarchical graph, in which four key variables (i.e., \({A}_{b}\), \({X}_{b}\), \({A}_{t}\), \({X}_{t}\)) need to be clarified.
(1) The adjacency matrix \({A}_{b}\in {\{{{{{\mathrm{0,1}}}}}\}}^{n\times n}\) in the protein graph and protein contact map are exactly equivalent. Contact maps are obtained with atomic level 3D coordinates of proteins. First, we retrieve the native protein structures from the Protein Data Bank^{60} and protein structures of various RMSD scores by AlphaFold^{43}. Then we represent the location of each residue by the 3D coordinate of its \({C}_{\alpha }\) atom. The presence or the absence of contact between a pair of residues is decided by their \({C}_{\alpha }{C}_{\alpha }\) physical distance. We perform a sensitivity analysis (see Supplementary Fig. 8) and find that our model produces similar results when trained on contact maps with cutoff distances ranging between 9 Å to 12 Å. Finally, we choose the optimal cutoff distance of 10 Å, which allows our model to peak its performance. (2) For a feature matrix \({X}_{b}\), each row represents a set of properties for one amino acid residue. In this work, seven residuelevel properties are considered (i.e., \(\theta=7\)): isoelectric point, polarity, acidity and alkalinity, hydrogen bond acceptor, hydrogen bond donor, octanolwater partition coefficient, and topological polar surface area. Supplementary Data File 3 contains quantitative values of seven types of properties for each amino acid. All properties can be easily retrieved from the RDKit repository^{61}. (3) The PPI network structure determines the adjacency matrix \({A}_{t}\in {\{{{{{\mathrm{0,1}}}}}\}}^{m\times m}\), in which the \(i\)th row and \(j\)th column element is 1 if the \(i\)th and \(j\)th proteins interact. (4) The \(i\)th row of the feature matrix \({X}_{t}\) represents the representation vector for the \(i\)th protein graph \({g}_{b}\).
BGNN for learning protein representations
We use the bottom view graph neural networks (BGNN) to learn protein representations. Graph convolutional networks (GCNs) have shown great effectiveness for relational data and are suitable for learning graphstructured protein representations. Thus, we propose BGNN based on GCNs.
Given the adjacency matrix \({A}_{b}\in {\{{{{{\mathrm{0,1}}}}}\}}^{n\times n}\) and the feature matrix \({X}_{b}\in {{\mathbb{R}}}^{n\times \theta }\) of an arbitrary protein graph \({g}_{b}\), BGNN outputs the residuelevel representations in the first GCN block, \({H}^{(1)}\in {{\mathbb{R}}}^{n\times {d}_{1}}\):
where \({d}_{1}\) is the embedding dimension for the first GCN layer.
Formally, we update residue representations with the neighbor aggregations based on the work of Kipf and Welling^{36}:
where \({I}_{n}\in {{\mathbb{R}}}^{n\times n}\) is the identity matrix, \(\widetilde{D}\in {{\mathbb{R}}}^{n\times n}\) is the diagonal degree matrix with entries \({D}_{{ii}}={\sum }_{j}{\left({A}_{b}+{I}_{n}\right)}_{{ij}}\), \({W}^{(1)}\in {{\mathbb{R}}}^{\theta \times {d}_{1}}\) is a learnable weight matrix for the GCN layer, ReLU, BN denotes the ReLU activation function and batch normalization, respectively.
With the learnable weight matrix \({W}^{(2)}\in {{\mathbb{R}}}^{{d}_{1}\times {d}_{2}}\), the second GCN block produces the output \({H}^{(2)}\in {{\mathbb{R}}}^{n\times {d}_{2}}\):
Finally, we perform the readout operation with a selfattention graph pooling layer^{39} and average aggregation to obtain the entire graph representation of a fixed size, \(x\in {{\mathbb{R}}}^{1\times {d}_{2}}\).To clarify, we use \({x}_{i}\in {{\mathbb{R}}}^{1\times {d}_{2}}\) to represent the final representation for the \(i\)th protein graph.
TGNN for learning PPI network information
We use the top view graph neural networks (TGNN) to learn PPI network information. We are inspired by graph isomorphism network (GIN^{37}), which has the superb expressive power to capture graph structures. Formally, we are given the PPI graph \({g}_{t}=({V}_{t},{A}_{t},{X}_{t})\), where \({X}_{t}\in {{\mathbb{R}}}^{m\times {d}_{2}}\) is defined as the feature matrix whose row vector is a final protein representation from BGNN (i.e., \({X}_{t}^{\left[i,:\right]}={x}_{i},i={{{{\mathrm{1,2}}}}},\ldots,m\)). TGNN updates the representation of protein \(v\) in the \(k\)th GIN block:
where \({x}_{v}^{(k)}\) denotes the representation of protein \(v\) after the \(k\)th GIN block, \({{{{{\mathscr{N}}}}}}({{{{{\mathcal{v}}}}}})\) is a set of proteins adjacent to \(v\), and \(\epsilon\) is a learnable parameter. We denote the inputs of protein representations for the first GIN block as \({x}_{i}^{(0)}={x}_{i},i={{{{\mathrm{1,2}}}}},\ldots,m\).
After three GIN blocks, TGNN produces representations for all proteins. For an arbitrary query pair containing the \(i\)th and \(j\)th proteins, we use the concatenation operation to combine the representations of \({x}_{i}^{(3)}\) and \({x}_{j}^{(3)}\). A fully connected layer (FC) is employed as the classifier. The final vector \({\hat{y}}_{{ij}}\in {{\mathbb{R}}}^{1\times c}\) for the presence probability of PPI is denoted as \({\hat{y}}_{{ij}}={{{{{\rm{FC}}}}}}\left({h}_{i}^{(3)}{}{h}_{j}^{(3)}\right)\) where \(c\) denotes the total number of PPI types involved and \(\parallel\) denotes the concatenation operation.
Model training details
Given a training set \({{{{{{\mathscr{X}}}}}}}_{{train}}\) and ground truth labels for multitype PPIs \({{{{{{\mathscr{Y}}}}}}}_{{train}}\), we train BGNN and TGNN in an endtoend manner by minimizing the loss function of multitask binary crossentropy:
where \(\Theta\) is the set of all learnable parameters, and \({ij}\) denotes the ground truth of the \(k\)th type PPI of the \(i\)th and \(j\)th proteins.
We determine all the hyperparameters through a grid search based on a 5fold crossvalidation. For BGNN, we set the output dimension \({d}_{1}\), \({d}_{2}\) of weight matrix to 128. For each GIN block in TGNN, we use a twolayer MLP and set the output dimension of each layer to 64. As the STRING dataset contains seven types of PPIs, we set the output dimension of the FC layer to \(c=7\). We use the Adam optimizer with a learning rate \({lr}=0.001\), \({\beta }_{1}=0.99\), \({\beta }_{2}=0.99\), a batch size of 128, and the default epoch number of 500. We train all of the model parameters until convergence in each crossvalidation.
Residue importance computation
We employ the method called GNNExplainer^{62} to generate explanations for HIGHPPI. By taking the welltrained GNN model and its predictions as inputs, GNNExplainer returns the most important subgraph by maximizing the mutual information \({MI}\) between the model prediction and possible subgraphs. Motivated by this, we directly formalize the notion of subgraph importance using \({MI}\) and further compute the importance of all nodes (i.e., residues).
Given protein graphs \({G}_{1}\) and \({G}_{2}\) that connect in the PPI network, our goal is to identify the node importance of \({G}_{1}\). According to GNNExplainer, once sampling a random subgraph \({G}_{s}\subseteq {G}_{1}\), we obtain the entire importance of \({G}_{s}\) as follow:
where \({{MI}}_{s}\) represents importance of \({G}_{s}\), \(Y\) is a variable indicating the probability of PPI presence of \({G}_{1}\) and \({G}_{2}\), and \(H\left(\bullet \right)\) is the entropy term.
Assume that all nodes in the subgraph \({G}_{s}\) contribute equally to the \({MI}\) value, we obtain the batch importance for each node in \({G}_{s}\). The final importance score for a specific node is the average of all its batch importance scores. For example, if a node \(v\) contributes 0.4 and 0.6 for two sampled subgraphs respectively, the final importance of node \(v\) is 0.5. To facilitate comparison, we compute the zscores of final residue importance for standardization:
where \({z}_{f}\in {{\mathbb{R}}}^{1\times n}\) is the finally computed importance vector for all residues, \(\mu\) is the average of \({z}_{f}\), \(\mu\) is the standard deviation of \({z}_{f}\), and \({z}_{s}\in {{\mathbb{R}}}^{1\times n}\) is the zscore importance after standardization.
Statistics and reproducibility
As indicated in figure legends, data in bar charts are represented as mean \(\pm\) standard deviation (SD). For all boxplots, the center line represents the median, upper and lower edges represent the interquartile range, and the whiskers represent 0.5× interquartile range. The statistical significance between the two groups was obtained by a twosided ttest with Pvalue < 0.05 considered significant.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Data availability
The PPI and protein data used in this study are available in the Zenodo database under “Accession Code 7213401”. They are obtained from the following publicly available database. Datasets containing protein sequences and their interaction annotations are obtained from https://github.com/muhaochen/seq_ppi. The native protein structures are obtained from PDB: https://www.rcsb.org/. Protein structures with errors are obtained from AlphaFold: https://alphafold.ebi.ac.uk/. The catalytic site information of proteins can be found at CSA: https://www.ebi.ac.uk/thorntonsrv/mcsa/. The ground truth of binding site information is obtained from PDBePISA: https://www.ebi.ac.uk/pdbe/pisa/. All other relevant data supporting the key findings of this study are available within the article and its Supplementary Information files or from the corresponding author upon reasonable request. Source data are provided with this paper.
Code availability
An opensource software implementation of HIGHPPI is available at https://github.com/zqgao22/HIGHPPI. The source code can be cited by using https://doi.org/10.5281/zenodo.7600622.
References
Petta, I. et al. Modulation of protein–protein interactions for the development of novel therapeutics. Mol. Ther. 24, 707–718 (2016).
Skrabanek, L., Saini, H. K., Bader, G. D. & Enright, A. J. Computational prediction of protein–protein interactions. Mol. Biotechnol. 38, 1–17 (2008).
Hope, K. J., Jin, L. & Dick, J. E. Acute myeloid leukemia originates from a hierarchy of leukemic stem cell classes that differ in selfrenewal capacity. Nat. Immunol. 5, 738–743 (2004).
Zeng, A. G. et al. A cellular hierarchy framework for understanding heterogeneity and predicting drug response in acute myeloid leukemia. Nat. Med. 28, 1–12 (2022).
Couturier, C. P. et al. Singlecell RNAseq reveals that glioblastoma recapitulates a normal neurodevelopmental hierarchy. Nat. Commun. 11, 1–19 (2020).
Engelberg, K., Bechtel, T., Michaud, C., Weerapana, E. & Gubbels, M. J. Proteomic characterization of the Toxoplasma gondii cytokinesis machinery portrays an expanded hierarchy of its assembly and function. Nat. Commun. 13, 1–15 (2022).
Wigbers, M. C. et al. A hierarchy of protein patterns robustly decodes cell shape information. Nat. Phys. 17, 578–584 (2021).
Ho, T. S. Y. et al. A hierarchy of ankyrinspectrin complexes clusters sodium channels at nodes of Ranvier. Nat. Neurosci. 17, 1664–1672 (2014).
Siegle, J. H. et al. Survey of spiking in the mouse visual system reveals functional hierarchy. Nature 592, 86–92 (2021).
Hendrikx, E., Paul, J. M., van Ackooij, M., van der Stoep, N. & Harvey, B. M. Visual timingtuned responses in human association cortices and response dynamics in early visual cortex. Nat. Commun. 13, 1–19 (2022).
Zhang, Y. et al. A system hierarchy for braininspired computing. Nature 586, 378–384 (2020).
Guharoy, M., Lazar, T., MacossayCastillo, M. & Tompa, P. Degron masking outlines degronons, codegrading functional modules in the proteome. Commun. Biol. 5, 1–15 (2022).
Wu, C. H. et al. Identification of lncRNA functions in lung cancer based on associated proteinprotein interaction modules. Sci. Rep. 6, 1–11 (2016).
Rolland, T. et al. A proteomescale map of the human interactome network. Cell 159, 1212–1226 (2014).
Hein, M. Y. et al. A human interactome in three quantitative dimensions organized by stoichiometries and abundances. Cell 163, 712–723 (2015).
Huttlin, E. L. et al. Architecture of the human interactome defines protein communities and disease networks. Nature 545, 505–509 (2017).
Kaboord, B. & Perr, M. Isolation of proteins and protein complexes by immunoprecipitation. Methods Mol. Biol. 424, 349–364 (2008).
Aronheim, A., Zandi, E., Hennemann, H., Elledge, S. J. & Karin, M. Isolation of an AP1 repressor by a novel method for detecting proteinprotein interactions. Mol. Cell. Biol. 17, 3094–3102 (1997).
Su, J. F., Huang, Z., Yuan, X. Y., Wang, X. Y. & Li, M. Structure and properties of carboxymethyl cellulose/soy protein isolate blend edible films crosslinked by Maillard reactions. Carbohydr. Polym. 79, 145–153 (2010).
Zhao, L., Wang, J., Hu, Y. & Cheng, L. Conjoint feature representation of GO and protein sequence for PPI prediction based on an inception RNN attention network. Mol. Ther.Nucleic Acids 22, 198–208 (2020).
Renaud, N. et al. DeepRank: a deep learning framework for data mining 3D proteinprotein interfaces. Nat. Commun. 12, 1–8 (2021).
Kov´acs, I. A. et al. Networkbased prediction of protein interactions. Nat. Commun. 10, 1–8 (2019).
Nasiri, E., Berahmand, K., Rostami, M. & Dabiri, M. A novel link prediction algorithm for proteinprotein interaction networks by attributed graph embedding. Computers Biol. Med. 137, 104772 (2021).
Lv, G., Hu, Z., Bi, Y. & Zhang, S. Learning unknown from correlations: graph neural network for internovelprotein interaction prediction. In 30th International Joint Conference on Artificial Intelligence (IJCAI). https://doi.org/10.48550/arXiv.2105.06709 (2021).
Kulmanov, M., Khan, M. A. & Hoehndorf, R. DeepGO: predicting protein functions from sequence and interactions using a deep ontologyaware classifier. Bioinformatics 34, 660–668 (2018).
Chen, M. et al. Multifaceted protein–protein interaction prediction based on Siamese residual RCNN. Bioinformatics 35, i305–i314 (2019).
Hsieh, Y. L., Chang, Y. C., Chang, N. W. & Hsu, W. L. In Proc. 8th International Joint Conference on Natural Language Processing Vol. 2 (Short Papers) 240–245 (Asian Federation of Natural Language Processing, 2017).
Saha, S. & Raghava, G. P. S. Prediction of continuous Bcell epitopes in an antigen using recurrent neural network. Proteins: Struct., Funct., Bioinforma. 65, 40–48 (2006).
Gligorijević, V. et al. Structurebased protein function prediction using graph convolutional networks. Nat. Commun. 12, 1–14 (2021).
Jiménez, J. et al. DeepSite: proteinbinding site predictor using 3Dconvolutional neural networks. Bioinformatics 33, 3036–3042 (2017).
Amidi, A. et al. EnzyNet: enzyme classification using 3D convolutional neural networks on spatial representation. PeerJ 6, e4750 (2018).
Tubiana, J., SchneidmanDuhovny, D. & Wolfson, H. J. ScanNet: An interpretable geometric deep learning model for structurebased protein binding site prediction. Nat. Methods 19, 1–10 (2022).
Goldberg, D. S. & Roth, F. P. Assessing experimentally derived interactions in a small world. Proc. Natl Acad. Sci. 100, 4372–4376 (2003).
Fouss, F., Pirotte, A., Renders, J. M. & Saerens, M. Randomwalk computation of similarities between nodes of a graph with application to collaborative recommendation. IEEE Trans. Knowl. Data Eng. 19, 355–369 (2007).
Tong, H., Faloutsos, C. & Pan, J. Y. 6th International Conference on Data Mining (ICDM) (IEEE, 2006).
Kipf, T. N. & Welling, M. Semisupervised classification with graph convolutional networks. In 5th International Conference on Learning Representations (ICLR). https://doi.org/10.48550/arXiv.1609.02907 (2017).
Xu, K., Hu, W., Leskovec, J. & Jegelka, S. How powerful are graph neural networks? In 7th International Conference on Learning Representations (ICLR). https://doi.org/10.48550/arXiv.1810.00826 (2019).
Szklarczyk, D. et al. STRING v11: protein–protein association networks with increased coverage, supporting functional discovery in genomewide experimental datasets. Nucleic acids Res. 47, D607–D613 (2019).
Lee, J., Lee, I. & Kang, J. Selfattention graph pooling. In 36th International Conference on Machine Learning (ICML). https://doi.org/10.48550/arXiv.1904.08082 (2019).
Zheng, S., Li, Y., Chen, S., Xu, J. & Yang, Y. Predicting drug–protein interaction using quasivisual question answering system. Nat. Mach. Intell. 2, 134–140 (2020).
Wong, L., You, Z. H., Li, S., Huang, Y. A. & Liu, G. Detection of protein–protein interactions from amino acid sequences using a rotation forest model with a novel PRLPQ descriptor. Adv. Intell. Syst. Comput. https://doi.org/10.1007/9783319220536_75 (2015).
Park, Y. & Marcotte, E. M. Flaws in evaluation schemes for pairinput computational predictions. Nat. methods 9, 1134–1136 (2012).
Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021).
Yang, C., Rangarajan, A. & Ranka, S. Visual explanations from deep 3D convolutional neural networks for Alzheimer’s disease classification. AMIA Annu. Symp. Proc. 2018, 1571–1580 (2018).
Ming, Y. et al. Understanding hidden memories of recurrent neural networks. In 2017 IEEE Conference on Visual Analytics Science and Technology (VAST). https://doi.org/10.48550/arXiv.1710.10777 (2017).
Fernandes, J. & Gattass, C. R. Topological polar surface area defines substrate transport by multidrug resistance associated protein 1 (MRP1/ABCC1). J. medicinal Chem. 52, 1214–1218 (2009).
Hu, Z., Ma, B., Wolfson, H. & Nussinov, R. Conservation of polar residues as hot spots at protein interfaces. Proteins: Struct., Funct., Bioinforma. 39, 331–342 (2000).
Young, L., Jernigan, R. & Covell, D. A role for surface hydrophobicity in proteinprotein recognition. Protein Sci. 3, 717–729 (1994).
Korn, A. P. & Burnett, R. M. Distribution and complementarity of hydropathy in mutisunit proteins. Proteins: Struct., Funct., Bioinforma. 9, 37–55 (1991).
Katz, L. A new status index derived from sociometric analysis. Psychometrika 18, 39–43 (1953).
Zhou, T., Lv, L. & Zhang, Y. C. Predicting missing links via local information. Eur. Phys. J. B 71, 623–630 (2009).
Barabási, A. L. & Albert, R. Emergence of scaling in random networks. Science 286, 509–512 (1999).
Jeh, G. & Widom, J. Proc. 8th International Conference on Knowledge Discovery and Data Mining (ACM, 2002).
De Meo. P., Ferrara E., Fiumara G. & Provetti A. Generalized louvain method for community detection in large networks. In 11th International Conference on Intelligent Systems Design and Applications (ISDA). https://doi.org/10.1109/ISDA.2011.6121636 (2011).
Krissinel, E. & Henrick, K. Inference of macromolecular assemblies from crystalline state. J. Mol. Biol. 372, 774–797 (2007).
Porter, C. T., Bartlett, G. J. & Thornton, J. M. The Catalytic Site Atlas: a resource of catalytic sites and residues identified in enzymes using structural data. Nucleic acids Res. 32, D129–D133 (2004).
Selvaraju, R. R. et al. Gradcam: Visual explanations from deep networks via gradientbased localization. In Proceedings of the 2017 IEEE International Conference on Computer Vision (ICCV). https://doi.org/10.1007/s11263019012287 (2017).
Li, J. et al. Semisupervised graph classification: a hierarchical graph perspective. In 2019 The World Wide Web Conference (WWW). https://doi.org/10.48550/arXiv.1904.05003 (2019).
Boutet, E., Lieberherr, D., Tognolli, M., Schneider, M. & Bairoch, A. UniProtKB/SwissProt 89–112 (Humana Press, 2007).
Berman, H. M. et al. The protein data bank. Nucleic Acids Res. 28, 235–242 (2000).
Landrum, G., Tosco, P. & Kelley, B. rdkit/rdkit: 2021_09_4 (Q3 2021) 351 Release. https://zenodo.org/record/5835217#.Y_JocB9Bzcs (2022).
Ying, Z., Bourgeois, D., You, J., Zitnik, M. & Leskovec, J. Gnnexplainer: Generating explanations for graph neural networks. In 33rd Advances in Neural Information Processing Systems (NeurIPS). https://doi.org/10.48550/arXiv.1903.03894 (2019).
Acknowledgements
The research of Li was supported by National Natural Science Foundation of China (Grant No. 62206067), Tencent AI Lab RhinoBird Focused Research Program RBFR2022008 and GuangzhouHKUST(GZ) Joint Funding Scheme. The research of Huang was supported by the National Natural Science Foundation of China (Grant No. 21825101).
Author information
Authors and Affiliations
Contributions
Z.G. and C.J. wrote the first draft of manuscript. J.L., Y.H., and L.L. revised the manuscript to the submitted version. J.L., Z.G., Y.H., and H.Y. conceived the study. Z.G. designed all the experiments and wrote the codebase of HIGHPPI. Z.G., J.Z., and X.J. conduct the benchmarks, and run all of the analysis. X.J. collected and preprocessed protein contact maps. Z.G., L.L., and P.Z. contributed to data analysis and model discussion. J.Z. conducted the figure design for overall framework. Z.G., C.J., J.Z., and X.J. completed the visualizations. J.L. and Y.H. supervised the research. All of the authors reviewed the manuscript and approved it for submission.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature Communications thanks Pufeng Du and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Source data
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Gao, Z., Jiang, C., Zhang, J. et al. Hierarchical graph learning for protein–protein interaction. Nat Commun 14, 1093 (2023). https://doi.org/10.1038/s41467023367361
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41467023367361
This article is cited by

GNNGLPPI: multicategory prediction of proteinprotein interactions using graph neural networks based on global graphs and local subgraphs
BMC Genomics (2024)

A variational expectationmaximization framework for balanced multiscale learning of protein and drug interactions
Nature Communications (2024)

Multilevel characterization of unknown protein sequences using hierarchical long short term memory model
Multimedia Tools and Applications (2024)

A performance prediction method for onsite chillers based on dynamic graph convolutional network enhanced by association rules
Building Simulation (2024)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.