HNSPPI: a hybrid computational model combing network and sequence information for predicting protein–protein interaction

Abstract

Most life activities in organisms are regulated through protein complexes, which are mainly controlled via Protein–Protein Interactions (PPIs). Discovering new interactions between proteins and revealing their biological functions are of great significance for understanding the molecular mechanisms of biological processes and identifying the potential targets in drug discovery. Current experimental methods only capture stable protein interactions, which lead to limited coverage. In addition, expensive cost and time consuming are also the obvious shortcomings. In recent years, various computational methods have been successfully developed for predicting PPIs based only on protein homology, primary sequences of protein or gene ontology information. Computational efficiency and data complexity are still the main bottlenecks for the algorithm generalization. In this study, we proposed a novel computational framework, HNSPPI, to predict PPIs. As a hybrid supervised learning model, HNSPPI comprehensively characterizes the intrinsic relationship between two proteins by integrating amino acid sequence information and connection properties of PPI network. The experimental results show that HNSPPI works very well on six benchmark datasets. Moreover, the comparison analysis proved that our model significantly outperforms other five existing algorithms. Finally, we used the HNSPPI model to explore the SARS-CoV-2-Human interaction system and found several potential regulations. In summary, HNSPPI is a promising model for predicting new protein interactions from known PPI data.

HGNNPIP: A Hybrid Graph Neural Network framework for Protein-protein Interaction Prediction

“…In previous works, computational scientists tried to access various public databases to obtain the interacting data (positive instances), which were mostly validated by experiments. To construct balanced datasets, they manually generated the same number of negative instances through subcellular localization 2,35,37 . However, these negative PPIs are only associated with a small portion of proteins (from 7.58% to 47.22%), indicating serious data distribution bias on most benchmarking datasets ( Table 1 ).…”

“…As mentioned in the section of “ Materials and Methods ”, most computational scientists tend to generate negative pairs by randomly pairing proteins with different subcellular localization 2,34,35,37 . Therefore, these representative benchmark PPI datasets are manually curated with an equal number of positive and negative samples.…”

“…Although subcellular localization 2 is the most commonly used strategy to generate negative instances, it remains unknown whether the constructed negative set will cause obvious false positives. In this study, we proposed three potential strategies for negative instance generation and determined the best one for PPI prediction.…”

“…The representative amino acid sequence-based models include DeepFE-PPI 28 , DPPI 30 , PIPR 29 , and DeepTiro 27 . Until recently, topology information of PPI networks began to be integrated with protein sequences for PPI computational identification (e.g., S-VGAE 31 ; HNSPPI 2 , etc .). To train the models with numerical data, various deep learning architectures were designed to implement feature embedding.…”

Sequence-based Protein-Protein Interaction Prediction Using Multi-kernel Deep Convolutional Neural Networks with Protein Language Model

“…TAGPPI (8) incorporates both sequence features and predicted structural information and employs graph representation learning methods on contact maps to obtain 3D structure features of proteins. HNSPPI (9) adopts a feature fusion strategy of both network topology and sequence information for comprehensive feature extraction and employs a simple classifier for predictions, making it both lightweight and efficient. Graph-BERT (10) utilizes a language model-based embedding SeqVec to represent protein sequences and a graph convolutional neural network with the training strategy of subgraph batches using a top-k intimacy sampling approach.…”