Improved prediction of post-translational modification crosstalk within proteins using DeepPCT

Area of Science:

• Computational biology
• Bioinformatics
• Molecular biology

Background:

• Post-translational modification (PTM) crosstalk is crucial for biological processes.
• Current machine learning methods for identifying PTM crosstalk lack sufficient accuracy.
• Advances in deep learning and protein structure prediction offer new avenues for PTM crosstalk identification.

Purpose of the Study:

• To develop a highly accurate deep learning algorithm for identifying PTM crosstalk.
• To leverage protein structure information for improved PTM crosstalk prediction.
• To enhance the generalizability of PTM crosstalk prediction models.

Main Methods:

• Developed DeepPCT, a deep learning algorithm integrating sequence and structure information.
• Employed AlphaFold2-predicted protein structures for structure-based prediction.
• Utilized cross-attention mechanisms for sequence-based prediction and graph neural networks for structure-based prediction.
• Integrated three classifiers: sequence-based deep learning, structure-based deep learning, and structure-based machine learning.

Main Results:

• DeepPCT significantly outperformed existing algorithms in identifying PTM crosstalk.
• The algorithm demonstrated superior generalizability on new datasets due to reduced distance dependency.
• Combined sequence and structure-based predictions to achieve robust performance.

Conclusions:

• DeepPCT represents a significant advancement in predicting PTM crosstalk.
• The integration of deep learning with protein structural data enhances prediction accuracy and generalizability.
• The developed algorithm provides a valuable tool for studying PTM crosstalk in biological systems.