Scholarlink

Protein–protein interactions (PPIs) are responsible for numerous essential physiological processes and molecular pathways signaling pathways and underlying disease mechanisms. Proteins participate in diverse biological processes within the cell by means of their interactions with other proteins. Protein–protein interactions are commonly represented as graphs, where proteins are nodes and interactions are binary edges. Numerous techniques have been suggested for the computational prediction of PPIs using various criteria for evaluating their performance. This paper, LSTM-PPI, proposes a novel deep convolution strategy utilizing the one-hot encoding and physicochemical properties of proteins for prediction of PPI. The performance is evaluated on the Human Protein dataset. Our proposed model demonstrated high accuracy, F1-score, AUC score, and AUPRC score. On these datasets, the suggested model obtains an 87% prediction accuracy score, an AUC score of 94.6%, and an AUPRC score of 97%. Our proposed model LSTM-PPI outperforms all the state-of-the-art methods.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

LSTM-PPI: A Deep Learning-Based Protein–Protein Interaction Prediction Strategies for Human Proteome

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Arghya Bera,
  • Soumyendu Sekhar Bandyopadhyay

摘要

Protein–protein interactions (PPIs) are responsible for numerous essential physiological processes and molecular pathways signaling pathways and underlying disease mechanisms. Proteins participate in diverse biological processes within the cell by means of their interactions with other proteins. Protein–protein interactions are commonly represented as graphs, where proteins are nodes and interactions are binary edges. Numerous techniques have been suggested for the computational prediction of PPIs using various criteria for evaluating their performance. This paper, LSTM-PPI, proposes a novel deep convolution strategy utilizing the one-hot encoding and physicochemical properties of proteins for prediction of PPI. The performance is evaluated on the Human Protein dataset. Our proposed model demonstrated high accuracy, F1-score, AUC score, and AUPRC score. On these datasets, the suggested model obtains an 87% prediction accuracy score, an AUC score of 94.6%, and an AUPRC score of 97%. Our proposed model LSTM-PPI outperforms all the state-of-the-art methods.