Long-distance dependency combined multi-hop graph neural networks for protein–protein interactions prediction

Abstract

Background Protein–protein interactions are widespread in biological systems and play an important role in cell biology. Since traditional laboratory-based methods have some drawbacks, such as time-consuming, money-consuming, etc., a large number of methods based on deep learning have emerged. However, these methods do not take into account the long-distance dependency information between each two amino acids in sequence. In addition, most existing models based on graph neural networks only aggregate the first-order neighbors in protein–protein interaction (PPI) network. Although multi-order neighbor information can be aggregated by increasing the number of layers of neural network, it is easy to cause over-fitting. So, it is necessary to design a network that can capture long distance dependency information between amino acids in the sequence and can directly capture multi-order neighbor information in protein–protein interaction network. Results In this study, we propose a multi-hop neural network (LDMGNN) model combining long distance dependency information to predict the multi-label protein–protein interactions. In the LDMGNN model, we design the protein amino acid sequence encoding (PAASE) module with the multi-head self-attention Transformer block to extract the features of amino acid sequences by calculating the interdependence between every two amino acids. And expand the receptive field in space by constructing a two-hop protein–protein interaction (THPPI) network. We combine PPI network and THPPI network with amino acid sequence features respectively, then input them into two identical GIN blocks at the same time to obtain two embeddings. Next, the two embeddings are fused and input to the classifier for predict multi-label protein–protein interactions. Compared with other state-of-the-art methods, LDMGNN shows the best performance on both the SHS27K and SHS148k datasets. Ablation experiments show that the PAASE module and the construction of THPPI network are feasible and effective. Conclusions In general terms, our proposed LDMGNN model has achieved satisfactory results in the prediction of multi-label protein–protein interactions.

Simplifying Subgraph Representation Learning for Scalable Link Prediction

“…Graph representation learning (GRL) has numerous applications in drug discovery [Xiong et al, 2019], knowledge graph completion , and recommender systems [Ying et al, 2018]. The key downstream tasks in GRL are node classification , link prediction [Zhang and ] and graph classification .…”

“…Recently, a new research direction has emerged in GNNs on scalable training and inference techniques, which mainly focuses on node classification tasks. A common practice is to use different sampling techniques-such as node sampling , layer-wise sampling [Zou et al, 2019], and graph sampling [Zeng et al, 2020]-to subsample the training data to reduce computation. Other approaches utilize simplification techniques, such as removing intermediate non-linearities [Wu et al, 2019;, to speed up the training and inference.…”

“…Graphs are ubiquitous in representing relationships between interacting entities in a variety of contexts, ranging from social networks to polypharmacy [Zitnik et al, 2018]. One of the main tasks on graphs is link prediction [Liben-Nowell and Kleinberg, 2003], which involves predicting future or missing relationships between pairs of entities, given the structure of the graph.…”

“…One of the main tasks on graphs is link prediction [Liben-Nowell and Kleinberg, 2003], which involves predicting future or missing relationships between pairs of entities, given the structure of the graph. Link prediction plays a fundamental role in impactful areas, including molecular interaction prediction [Huang et al, 2020], recommender systems [Ying et al, 2018], and protein-protein interaction prediction [Zhong et al, 2022].…”

“…The core issue hampering the deployment of GNNs and SGRLs is their high computational demands on large-scale graphs. Many approaches have been proposed for GNNs to relieve this bottleneck, ranging from sampling Zeng et al, 2020] to simplification [Wu et al, 2019] techniques. However, these techniques fail to be applied directly in SGRL models, where the issue is even more adverse due to subgraph extractions pertaining to each link.…”