DeepPhosPPI: a deep learning framework with attention-CNN and transformer for predicting phosphorylation effects on...

Area of Science:

• Biochemistry
• Computational Biology
• Bioinformatics

Background:

• Protein phosphorylation is crucial for cellular signaling and disease, regulating protein activity and interactions.
• Understanding phosphorylation's impact on protein-protein interactions (PPIs) is vital for disease research.
• Current experimental validation methods are costly and time-consuming.

Purpose of the Study:

• To develop an efficient computational framework, DeepPhosPPI, for predicting phosphorylation's effects on PPIs.
• To leverage deep learning and protein language models for accurate prediction of phosphorylation-mediated PPI modulation.

Main Methods:

• Developed DeepPhosPPI, a sequence-based deep learning framework.
• Utilized pre-trained protein language models (ProtBERT, ESM-2) for feature embedding.
• Integrated attention-based convolutional neural networks and Transformer models.

Main Results:

• DeepPhosPPI accurately predicts phosphorylation effects on PPIs.
• The framework consistently outperformed state-of-the-art methods in benchmark tasks.
• Achieved high accuracy in functional site identification and regulatory effect classification.

Conclusions:

• DeepPhosPPI offers an efficient and accurate computational solution for predicting phosphorylation effects on PPIs.
• This framework can accelerate research in phosphorylation-dependent biological processes and related diseases.
• DeepPhosPPI advances the field of bioinformatics by integrating advanced deep learning techniques.