CBSG-PPI:基于图神经网络的蛋白质-蛋白质相互作用的预测算法

CBSG-PPI:A Prediction Algorithm for Protein-protein Interaction Based on Graph Neural Networks

为提高蛋白质-蛋白质相互作用(protein-protein interaction,PPI)预测的准确性,并深入探索细胞信号传导和疾病发生的生物学机制,本文提出一种简称为CBSG-PPI的预测算法。该算法首先利用3层前馈网络来处理蛋白质的k-mer特征,采用CT方法和Bert方法提取蛋白质的氨基酸序列以及使用卷积神经网络提取蛋白质的序列特征,再结合图神经网络和多层感知机来准确预测PPI。与现有的预测技术相比,CBSG-PPI在准确率、F1分数、召回率和精确率等多个关键性能指标上展现了明显的优势,在公开数据集上分别达到了0.855、0.853、0.840和0.866的高分。此外,本算法采用了一种改进的参数调整方法,显著提高了计算效率,其预测速度比传统算法快了约140倍。这一显著的性能提升,不仅证实了CBSG-PPI在预测PPI方面的研究价值,也为未来蛋白质间相互作用网络的构建和分析提供了有用的计算工具。

To enhance the accuracy of predicting protein-protein interaction(PPI)and to further explore the biological mechanisms behind cellular signaling and disease onset,this paper introduced an prediction algorithm abbreviated as CBSG-PPI.Firstly the algorithm processed the k-mer features of proteins using a three-layer feedforward network,and employed the CT method and the Bert method to extract amino acid sequences of proteins,and utilized a convolutional neural network to extract sequence features of proteins.Then it combined a graph neural network and a multilayer perceptron to accurately predict PPI.Compared to existing prediction techniques,CBSG-PPI had shown a clear advantage on public datasets across several key performance metrics such as accuracy,F1 score,recall,and precision,achieving high scores of 0.855,0.853,0.840,and 0.866,respectively.Moreover,the algorithm adopted an improved method for parameter tuning,significantly enhancing computational efficiency,with prediction speeds approximately 140 times faster than traditional algorithms.This substantial improvement in performance not only proves the research value of CBSG-PPI in predicting PPI but also provides a powerful computational tool for the future construction and analysis of protein interaction networks.